1) From Search to AI: SEO vs. GEO

By 2030 AI search could increase to 65% of global search traffic. Everything we know about online brand discovery is starting to break.

People aren’t browsing pages anymore. They’re asking AI.

And that means the entire discovery layer of the internet is being rebuilt.

Users are no longer starting with a Google search, clicking on links, and browsing websites.

Increasingly, they’re turning to AI tools like ChatGPT, Claude or Perplexity to get direct answers, summaries, and recommendations.

No links. No websites. No page-one rankings.

Traditionally, SEO has been about helping businesses appear higher in Google results.

That meant optimizing websites to match search terms, earn backlinks, load quickly, and convert once the user arrived.

But that model depends on one thing: people clicking on search results.

𝗔𝗜 𝗿𝗲𝗺𝗼𝘃𝗲𝘀 𝘁𝗵𝗲 𝗰𝗹𝗶𝗰𝗸.

When users ask ChatGPT for help:

• “compare project management tools “

• “what’s the best CRM for startups”

• “find me a cheaper alternative to X”

They are not browsing. They are expecting a direct, summarized answer in the chat itself.

𝗪𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝗺𝗲𝗮𝗻?

• The old e-commerce model is under pressure

• Decisions increasingly happen before a website is ever visited - if it is visited at all.

• In this model, there is no clear click path, no landing page, no funnel in the traditional sense.

• Most performance metrics (traffic, CTR, bounce rate) are becoming irrelevant.

𝗧𝗵𝗲 𝘀𝗵𝗶𝗳𝘁 𝗳𝗿𝗼𝗺 𝗦𝗘𝗢 𝘁𝗼 𝗚𝗘𝗢:

We are moving from SEO (Search Engine Optimization) to GEO (Generative Engine Optimization):

• SEO was about ranking on a page → GEO is about being included in the answer

• SEO optimized for keywords and clicks → GEO optimizes for selection by AI

• SEO rewarded visibility in a list → GEO rewards presence in a decision

• SEO drove traffic after discovery → GEO influences discovery itself

• SEO competed for attention → GEO competes for inclusion

Traditional search was about helping users compare options.

AI is about delivering a single answer.

That means a big shift in business models and monetization will follow:

• Focus is moving from driving traffic to influencing decisions directly.

• Businesses will shift from controlling customer journeys to ensuring they are the chosen option.

• Less revenue will happen on websites and more at the moment of AI recommendation.

• Advertising will change from buying exposure to influencing AI’s final answer.

• Companies will lose direct influence over customers as AI platforms become the gatekeepers (of which brands customers see).

• Traditional KPIs will be replaced by new KPIs like AI inclusion rates, recommendation frequency or how often a brand is chosen in AI-generated answers.

𝗧𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝗿 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

(How) are you preparing for the shift?

Opinions: Panagiotis Kriaris, Graphic sources: Panagiotis Kriaris, ARK Invest

Subscribe now

2) The AI-Native Banking OS

It was just a matter of time. The AI-native banking operating system is here.

More than 60% of organisations expect to deploy agents within the next two years (Gartner).

𝗙𝗼𝗿 𝗯𝗮𝗻𝗸𝘀, 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗮 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹 𝘀𝗵𝗶𝗳𝘁:

• On the front-end, agents are inserting a new layer between banks and their customers. One that makes decisions.

• On the back-end, agents will do the same - sitting between systems and making decisions that used to be embedded in workflows and executed by people.

𝗧𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺?

Banks are not ready for either.

Most banks today run on dozens of systems - core banking, CRM, payments, cards, KYC, fraud – that don’t talk to each other. Serving a customer means switching between them, with an estimated 60% of the work spent fixing and coordinating across systems.

𝗦𝗼, 𝗵𝗼𝘄 𝗱𝗼 𝘆𝗼𝘂 𝘀𝗼𝗹𝘃𝗲 𝗳𝗼𝗿 𝘁𝗵𝗮𝘁?

By bringing channels, front office, and operations into one operating model - with shared context, common rules, and one execution layer where customers, employees, and AI agents work together.

Backbase calls this the Unified Frontline.

And it’s not a product. It is a new operating model.

They have basically invented a new category to describe what a bank looks like when its entire customer-facing operation runs as one system:

• Digital channels, front office and operations run on the same data, workflows, and policies.

• AI agents are treated as participants - with defined roles, governed authority, and full accountability.

The product is the outcome. The AI-native Banking OS. It is Backbase’s approach to get banks to the Unified Frontline.

𝗧𝗵𝗲 𝗕𝗮𝗻𝗸𝗶𝗻𝗴 𝗢𝗦 𝗱𝗼𝗲𝘀 𝗳𝗼𝘂𝗿 𝘁𝗵𝗶𝗻𝗴𝘀:

• Understand: it builds a unified picture of every customer, every account, and every case. This creates one shared source of truth across the entire frontline.

• Run: it executes workflows across employees, AI agents, and banking systems. Both structured, repeatable processes and adaptive agentic tasks run through the same execution layer.

• Authorize: every action - by a human, a workflow, or an AI agent - must be authorized before it executes. The Banking OS governs who can do what, under what conditions, with what level of autonomy. Every decision is traceable and auditable.

• Optimize: the system learns from every execution. AI models improve. Decisions get sharper. Operations become more efficient over time.

𝗪𝗵𝘆 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝗺𝗮𝘁𝘁𝗲𝗿?

Because it solves banks’ core structural weakness - the fragmentation described above - and by doing so it enables them to compete in the agentic economy.

Banks can now scale without having to increase their headcount in a linear way. Agents will work alongside people to handle more customer requests, automate tasks, personalize experiences, and advise clients.

Banking re-invented.

Opinions: Panagiotis Kriaris, Graphic source: Backbase

Subscribe now

3) The Digital Identity Model

It is identity’s next phase. Just as payment networks connect banks globally, identity networks will connect financial institutions, businesses, and ID schemes into a single layer.

One integration, global access to verified identity.

𝗛𝗼𝘄 𝗽𝗮𝘆𝗺𝗲𝗻𝘁 𝗻𝗲𝘁𝘄𝗼𝗿𝗸𝘀 𝗰𝗼𝗻𝗻𝗲𝗰𝘁 𝗯𝗮𝗻𝗸𝘀:

• A shared network sits in the middle: schemes like Visa and Mastercard connect participants through a common rulebook

• Four globally standardized roles: issuer (cardholder’s bank), acquirer (merchant’s bank), merchant, customer

• One global integration: a bank connects once to the network and can transact with thousands of banks worldwide without bilateral agreements

• Trust is outsourced to the network: authentication, authorization, dispute rules, and settlement frameworks are standardized across markets

• Interoperability by design: it doesn’t matter where the card is issued or where it’s used - the network makes it work seamlessly

𝗛𝗼𝘄 𝗶𝗱𝗲𝗻𝘁𝗶𝘁𝘆 𝗻𝗲𝘁𝘄𝗼𝗿𝗸𝘀 𝘄𝗼𝗿𝗸 – the parallel:

• Bring multiple government eID schemes into a single network interface, replacing country-by-country integrations

• Standardize trust: map and align different assurance levels across schemes into a common, comparable framework

• Act as a translation layer: handle differences in formats, protocols, and regulatory requirements across identity systems

• Enable selective disclosure: support attribute-level data exchange (e.g. age, residency, eligibility) instead of full identity datasets

• Allow verified credentials to be used across multiple services and interactions

• Incorporate local regulatory requirements into how identity data is accessed and shared

• Integrate verified, government-backed identity directly into transaction and onboarding flows

• Provide a single access point to multiple identity providers

• Enable new transaction models: support identity-dependent flows such as regulated purchases, credit, and age-restricted services

𝗧𝗵𝗲 𝗶𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝘁𝗵𝗶𝗻𝗴 𝗶𝘀 𝘁𝗵𝗮𝘁 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗶𝗻 𝗽𝗹𝗮𝗰𝗲 𝘁𝗼𝗱𝗮𝘆.

Companies like Hopae sit between institutions and national eID schemes as plug-and-play infrastructure, removing the need to integrate and manage each system separately.

Built for cross-border use from day one and designed around how regulated institutions actually operate: the same integration models, compliance requirements, and liability structures.

The framing is intentional and reflects the fact that identity is a network problem. Because it only works when multiple parties can rely on the same verified information.

𝗧𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝗿 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

For institutions and players relying on identity verification, the question is no longer whether identity networks will replace separate integrations, but how big the gap is to operating this way.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

4) The Agentic Organisation

We are moving from the digital to the agentic organization: automation replaced tasks - now agents will run entire workflows.

But there is one misunderstanding. An agentic organization is not an organization without humans but one where humans shift roles - moving from execution to oversight, judgment, and creativity, while AI agents run the workflows.

McKinsey has analysed the biggest changes. Here are my top take-aways:

• 𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝗶𝗼𝗻𝘀 𝗺𝗼𝘃𝗲 𝘁𝗼 𝗔𝗜-𝗻𝗮𝘁𝗶𝘃𝗲 𝗰𝗵𝗮𝗻𝗻𝗲𝗹𝘀. Instead of apps and websites, every customer could have a personal AI concierge - negotiating, learning, and adapting in real time. Companies that own these interfaces, and the proprietary data behind them, will hold the competitive edge.

• 𝗘𝗻𝘁𝗶𝗿𝗲 𝗲𝗻𝗱-𝘁𝗼-𝗲𝗻𝗱 𝗷𝗼𝘂𝗿𝗻𝗲𝘆𝘀 𝗿𝘂𝗻 𝗯𝘆 𝗮𝗴𝗲𝗻𝘁𝘀. Small human teams will supervise dozens of agents - two or three people could guide 50–100 agents handling everything from customer onboarding to compliance checks.

• 𝗙𝗿𝗼𝗺 𝘀𝗶𝗹𝗼𝘀 𝘁𝗼 𝗻𝗲𝘁𝘄𝗼𝗿𝗸𝘀. Hierarchies and even agile teams will feel slow compared to flat, outcome-focused agentic teams. Instead of rigid org charts, we’ll see “work charts” where tasks and outcomes flow dynamically between humans and agents.

• 𝗥𝗲𝗮𝗹-𝘁𝗶𝗺𝗲, 𝗲𝗺𝗯𝗲𝗱𝗱𝗲𝗱 𝗴𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲. Control agents will enforce guardrails, compliance agents will monitor regulation, and critic agents will challenge outputs. Humans still hold accountability - but they step in only where oversight or judgment is required.

• 𝗧𝗵𝗲 𝗻𝗮𝘁𝘂𝗿𝗲 𝗼𝗳 𝘄𝗼𝗿𝗸 𝘄𝗶𝗹𝗹 𝘀𝗵𝗶𝗳𝘁. Some people will act as generalists, orchestrating agents across domains. Others will apply deep expertise to reimagine processes and handle exceptions. And many will focus on frontline interactions, spending less time on systems and more time with people. Culture and trust will be the glue that makes this hybrid model work.

• 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗯𝘂𝗶𝗹𝘁 𝗮𝗿𝗼𝘂𝗻𝗱 𝗮𝗴𝗲𝗻𝘁𝘀. Instead of heavy, monolithic IT, organizations will use interoperable agents that can plug into each other, into machines, and even into legacy platforms. Because AI evolves so fast data and vendor workflows will need to be kept separate, using open standards where possible, and ensuring agents can be swapped in and out as better tools emerge.

• 𝗧𝗵𝗲 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 𝗶𝘀 𝘀𝗽𝗲𝗲𝗱. AI systems move exponentially, whereas most organizations evolve linearly, new structures, new skills, new governance typically take years to unfold. Closing that gap means rethinking how fast organizations can adapt, and how much risk they are willing to take to keep up.

In the digital era, software connected people. In the agentic era, AI will connects agents, while humans operate on a layer above.

Adapted from McKinsey - The agentic organization: Contours of the next paradigm for the AI era.

Subscribe now

1) The AI Platform Bank

AI is rewriting the platform banking model.

And BNY is now showing what the next version looks like.

Platform banking has been mainly a distribution shift. It reflects a rethinking of how banks deliver services, driven by the platform economy.

The core shift was:

• From banks as destinations, where customers accessed bundled products through through bank channels

• To banks as providers of capabilities, embedded into third-party platforms and workflows

The shift allowed other players to build on top of bank infrastructure:

• Fintechs offering financial products without being banks

• Software platforms embedding payments or accounts into their workflows

• Corporates integrating banking directly into treasury and operations

BNY was an early example of this model, exposing its core infrastructure - custody, clearing, payments, and asset servicing—for direct integration into client workflows.

However, the model has one core weakness: it solves for distribution but not execution. It makes services accessible, but leaves clients do the actual work (to connect them, interpret the data, etc).

And this is where AI comes in.

• AI can aggregate and interpret data in real time

• It connects services into end-to-end workflows

• It turns data into actionable insights

• It can trigger and execute actions across services

BNY’s strategy is to apply AI across its platform to move from providing access to services to enabling execution within client workflows.

Example:

In the platform model, a treasury team would use BNY’s systems to access data and services -pull cash positions, analyse liquidity, and then initiate transfers across accounts.

With AI on top, these steps are no longer separate. The platform aggregates positions, interprets liquidity needs, and executes funding actions.

What it means:

• BNY already processes large volumes of financial data and sits inside critical institutional workflows. If it applies AI effectively, it can move up the value chain.

• Instead of processing transactions, BNY can coordinate and execute financial decisions across institutions. It moves from access to services to controlling how they are used in practice.

• This reduces client complexity, embeds the bank deeper into core workflows, and makes it harder to replace. It is a different position altogether where the bank is not just part of the process, but actively involved in running it.

This is a familiar AI pattern: BNY wants to turn platform banking into programmable infrastructure, where AI controls how services are used in practice. And BNY benefits by capturing more volume and higher-value activity.

Is this the future of platform banking?

Opinions: Panagiotis Kriaris, Graphic source: BNY

Subscribe now

2) Banking΄s Mythos Moment

Banking is going through its Mythos moment. But not in the sense that the headline might imply.

The launch of Anthropic’s Mythos might be posing one of the most significant structural challenges the industry has faced.

𝗧𝗵𝗲 𝗯𝗮𝗰𝗸𝗴𝗿𝗼𝘂𝗻𝗱:

• Claude Mythos, announced April 7 2026, is Anthropic’s most powerful AI model ever. But unlike previous models, they chose not to release it publicly for now.

• The reason is that it autonomously finds and exploits vulnerabilities across every major operating system and browser - with zero human guidance.

• It uncovered thousands of previously unknown security weaknesses, including flaws in critical software that had gone undetected for 27 years.

• Using Anthropic’s own words: the fallout if misused could be severe for economies, public safety and national security.

• Instead of a public launch, they created Project Glasswing - a $100mn controlled rollout to ~40 vetted partners: AWS, Apple, Microsoft, Google, JPMorgan Chase, NVIDIA, etc.

• UK banks are among the first financial institutions to get access, as part of a tightly controlled rollout aligned with regulators.

𝗧𝗵𝗲 𝗰𝗼𝗻𝗰𝗲𝗿𝗻𝘀:

The financial sector is uniquely exposed because of how it is built.

• Banking runs on deeply layered infrastructure - core systems, clearing networks, and multiple external dependencies - most of it not designed for continuous, machine-driven testing.

• Mythos introduces autonomous capability: it reads code, identifies weaknesses, tests them, and delivers working exploits without human involvement.

• This compresses the time between vulnerability discovery and exploitation from weeks or months to nreal-time.

• Critical financial infrastructure operates as a network. Weaknesses in clearing, settlement, or identity layers can spread across institutions and markets.

• Risk management frameworks are calibrated to periodic threats and response cycles but not to constant probing and adaptation.

• A small group of institutions will be able to find and fix vulnerabilities faster than others - but if misused, the same capability can be used to exploit them at scale.

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• Leading central banks and financial institutions are already treating this as a priority. Closed-door coordination is happening now.

• Supervisory models are being forced to adapt in real time. Existing approaches were not built for autonomous, continuously evolving threat environments.

• The paradox: the only way to defend against Mythos is to use Mythos.

• Cybersecurity will be moving out of compliance and into core strategy - directly linked to financial stability and systemic risk.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Is each bank alone in a position to defend itself against such a threat? And if not, what does a system-level response look like - and how realistic is it?

Opinions: Panagiotis Kriaris, Graphic source: Anthropic

Subscribe now

3) The Card Payments Stack

Card payments are a complex business. A layered stack of providers, orchestration, and schemes sits behind what externally looks like a simple click (or tap). Here’s how it works.

𝗧𝗵𝗲 𝗹𝗮𝘆𝗲𝗿𝘀:

1) Initiation

A customer taps or clicks pay - that’s all they see. From there, a multi-layered system determines approval, cost, and processing.

2) Acceptance

The payment is captured through POS systems, gateways, or PSPs. This layer determines which payment methods are available, how the checkout is structured, and how the transaction is passed into the network. It’s highly visible, but increasingly standardized.

3) Orchestration

Before the transaction reaches the network, it is directed across different acquirers and payment paths, checked against fraud and risk rules, and retried if needed. These choices are defined by the merchant’s setup and the providers they use.

4) Execution

Only at this stage does the transaction reach the core infrastructure: acquirers, card schemes, and issuing banks. This is where authorization, clearing, and settlement take place under strict scheme rules. It enables global acceptance and consistent processing across markets, but it is highly standardized, leaving limited room for differentiation.

𝗧𝗵𝗲 𝘁𝗿𝗲𝗻𝗱𝘀:

• Transaction outcomes are increasingly determined before they reach the schemes -through acquirer selection, routing rules, retry logic, and risk configuration applied at the merchant or PSP level.

• Payment acceptance is broadly standardized - most providers support the same methods and integrations, so it no longer creates a meaningful advantage on its own.

• The way merchants structure their payment setup now directly impacts results -combining providers, routing transactions, and setting retry strategies affects approval rates, cost, and failures.

• Card schemes continue to define the rules, handle authorization, and enable settlement - but an increasing share of how transactions are handled (e.g. routing, retries, risk checks) is determined earlier, by merchants and their providers.

• Approval rates, cost, and failure handling are no longer fixed characteristics of a single provider - they vary based on how transactions are distributed, managed, and continuously adjusted across the stack.

𝗧𝗵𝗲 𝗿𝗼𝗹𝗲 𝗼𝗳 𝗔𝗜:

• AI is increasing how often and how precisely transactions are adjusted after they are captured - routing, retries, and risk settings are updated continuously based on actual performance.

• This means the same providers no longer deliver the same results by default. Approval rates and costs now depend much more on how transactions are managed in real time than on the underlying infrastructure itself.

Opinions: Panagiotis Kriaris, Graphic source: twenty second

Subscribe now

4) B2B Payments’ missing layer

It’s an oxymoron. While everyone is focused on agentic commerce, payments have a (much simpler) $3 trillion problem waiting to be solved.

Now, there’s a contender.

On any given day $𝟯 𝘁𝗿𝗶𝗹𝗹𝗶𝗼𝗻 sits in accounts receivable - unpaid invoices companies are owed after delivering goods or services.

• Money earned but not yet collected

• Often tied up for weeks or months

• Requiring ongoing follow-ups, tracking, and reconciliation to convert into cash

That amount alone exceeds the size of the entire UK economy - locked in unpaid invoices. And this is just the US.

It is remarkable how, in the age of AI, a typical accounts receivable process is stuck in the past:

• Invoices, payments, and reconciliation sit in separate systems that don’t connect

• Payments arrive without context, requiring manual matching to invoices

• Finance teams spend time chasing updates instead of managing cash

• No real-time view of what’s been paid, pending, or overdue

• The process still relies on emails, spreadsheets, and human coordination

𝗔𝗻𝗱 𝗵𝗲𝗿𝗲 𝗰𝗼𝗺𝗲𝘀 𝘁𝗵𝗲 𝗱𝗶𝘀𝗿𝘂𝗽𝘁𝗶𝗼𝗻 𝗽𝗼𝗶𝗻𝘁.

A team of just 10 people – a New York–based fintech called Monk - has rebuilt this flow from contract to cash:

• Starts from the contract, structuring billing terms upfront instead of relying on manual invoice setup

• Matches payments to invoices automatically, even when data is incomplete or inconsistent

• Adapts follow-ups and actions based on actual payment behavior, not fixed reminder schedules

• Keeps invoicing, collections, and reconciliation in a single continuous flow instead of separate systems

• Handles exceptions within the system, reducing reliance on emails, spreadsheets, and manual work

𝗧𝗵𝗲 𝗿𝗲𝘀𝘂𝗹𝘁𝘀 𝗮𝗿𝗲 𝗶𝗺𝗽𝗿𝗲𝘀𝘀𝗶𝘃𝗲:

• 90%+ accuracy in turning contracts into invoices. Automatically

• 18 hours reported time savings per week. Time earned back to focus on growing the business

• 24% higher collection response rates

• +37% average increase in cash on hand

• 60% decrease in unpaid invoices

• 40%+ reduction in time to get paid

𝗧𝗵𝗶𝘀 𝗶𝘀 𝗮 𝗯𝗿𝗼𝗮𝗱𝗲𝗿 𝘀𝗵𝗶𝗳𝘁.

Because the problem is not unique to accounts receivable - it’s a pattern across financial workflows.

We have built systems around transactions, but have left the underlying logic - contracts, terms, context - outside, to be handled manually.

Now AI is bringing that logic into the flow, connecting it directly to action points and releasing huge amounts of time, cash, and operational capacity.

Monk has just announced their series A.

Watch out for companies solving large, existing problems while everyone else is focused on what might come next.

Opinions: Panagiotis Kriaris, Graphic source: Monk

Subscribe now

It was a matter of time. Amex announced its own agentic commerce play. Next to that of Visa and Mastercard. It is the agentic war of the schemes.

American Express:

• Announced ACE, a developer kit for agentic commerce, built around agent registration, account linking, intent capture, tokenized credentials, and cart context. It is not just a payments API but a full set of controls for agent-led transactions.

• Combines authorization, intent, merchant context, and membership value in one system, giving Amex more influence than networks that only see the payment step.

• Extends its trust model into agentic commerce: verified agents, explicit intent, spend controls, and protection for eligible purchases, reducing friction in delegated transactions.

• Brings offers, benefits, and merchant relationships directly into the agent flow, shaping which options are surfaced and selected.

• The strategic play is to move into the selection layer, where agents decide what gets bought and where value is captured

Visa:

• Visa’s agentic play is Visa Intelligent Commerce / Intelligent Commerce Connect, a merchant and developer layer for agent-initiated payments that supports agent protocols, works with token vaults, and integrates via APIs and MCP-based workflows.

• Extends its core stack (tokenization, network credentials, authentication) so agents can initiate transactions while authorization remains tied to Visa infrastructure.

• Makes payments programmable and callable by agents, without requiring Visa to control the interface or the agent itself.

• Positions itself as foundational infrastructure across merchant and developer environments, as shown through partnerships such as AWS.

• The strategic play is to ensure that, regardless of how transactions are initiated, execution still depends on Visa rails at scale.

Mastercard:

• Mastercard’s play is Agent Pay, combining Agentic Tokens with partnerships across AI platforms, orchestration providers, acquirers, and checkout players.

• Extends payment capabilities with a trust framework: registered agents, traceable tokens, and defined accountability across the transaction lifecycle.

• Embeds identity, authentication, and fraud signals into agent flows to validate the acting entity and its authority before execution.

• Expands its role beyond payment validation to include authorization of actors and actions - beyond transactions.

• The strategic play is to establish Mastercard as the layer that governs trusted, permissioned agent activity, positioning it deeper than pure execution.

Why:

Agentic commerce is forcing schemes to move beyond payments, as value is shifting from processing transactions to the steps before: selection, authorization, and execution.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) The AI Transformation Manifesto

This is a must read. The AI transformation manifesto.

Recently published by McKinsey. This is my summary.

1. Advantage from capabilities vs. technology

• Early AI winners are the same companies that were already strong operators

• Their edge is the ability to repeatedly absorb and apply new technology

• Capabilities compound over time, while one-off solutions don’t

2. A few areas drive most of the value

• A small number of areas disproportionately shape business economics

• Leaders focus there, instead of spreading across many use cases

• Treating all use cases equally dilutes impact

3. There is no value without business impact

• Leading companies focus on changes that materially shift outcomes

• Reinventing a few core domains beats optimizing many edges

• Marginal impact usually signals lack of focus

4. Senior business leaders should be in the driver’s seat

• Transformations are owned by business leaders, not delegated to IT

• Leadership needs both domain and technology understanding

• IT supports, but the business drives outcomes

5. AI transformation is a people transformation

• High talent density matters more than team size

• More capability moves in-house, especially in engineering

• Roles shift toward higher-value work and clearer ownership

6. Speed is the defining organizational advantage

• The key difference is how fast companies turn insight into action

• Faster allocation and fewer dependencies matter more than ideas

• Slower organizations rarely catch up

7. Tech platforms as strategic assets

• Platforms enable reuse, scale, and lower unit cost

• They determine how quickly teams can build and deploy

• Weak platforms slow execution across the board

8. Make data usable and valuable

• Data is often available but hard to use in practice

• Making it accessible and usable is the first step

• Advantage comes from improving quality and context over time

9. Design for adoption and build for scale

• Value is lost when solutions aren’t embedded into real workflows

• Adoption requires changes beyond the model itself

• Scaling needs to be designed upfront, not retrofitted

10. AI scale depends on trust

• Adoption depends on confidence from customers, regulators, and employees

• Failures quickly limit how far systems can be deployed

• Trust must grow alongside capability, especially with more autonomy

11. Agentic engineering as a crucial skill

• AI is moving from tools to systems that execute tasks end-to-end

• This requires new approaches to workflows, controls, and system design

• Early gains are already visible in areas like software development

12. Learning speed drives advantage

• Skills and knowledge become outdated faster than before

• Organizations need to continuously update how they operate

• Advantage comes from turning learning into action quickly

Summary: Panagiotis Kriaris, Source: McKinsey & Company

Subscribe now

3) NFC Payment Models

It is the technology behind contactless transactions. Mobile wallets, cards, smartwatches, phones as terminals, public transport gates. And the best part: it does not even touch the money.

NFC stands for Near-Field Communication. It is a wireless technology that lets two devices exchange payment information when they are close.

• NFC is a small chip built into devices like phones, cards, and smartwatches

• It works only at very short distance, which is why tap feels instant and secure

• It can connect different devices

• NFC doesn’t process the payment itself - it just passes the information to the system that actually moves the money (e.g. bank, card network, payment app)

However, NFC payment models are not one thing - they are built from three dimensions: who starts the payment, how the data is created, and whether the transaction is online or offline.

𝟭) 𝗪𝗵𝗼 𝘀𝘁𝗮𝗿𝘁𝘀 𝘁𝗵𝗲 𝗽𝗮𝘆𝗺𝗲𝗻𝘁:

• Payer-initiated

The customer device initiates the interaction and provides the payment details

Example: tapping a phone to another phone or device to send money

• Payee-initiated

The merchant initiates the transaction with a predefined amount

Example: a restaurant terminal presenting the bill for the customer to tap and pay

𝟮) 𝗛𝗼𝘄 𝘁𝗵𝗲 𝗽𝗮𝘆𝗺𝗲𝗻𝘁 𝗱𝗮𝘁𝗮 𝗶𝘀 𝗰𝗿𝗲𝗮𝘁𝗲𝗱:

• Static

The same payment information is reused every time and only transaction-specific elements change

Example: A taxi driver’s payment terminal always contains the same account details, and only the fare changes for each ride

• Dynamic

The payment data is generated per transaction and already includes the amount and context

Example: a ticket machine generating a specific fare for a single purchase

𝟯) 𝗖𝗼𝗻𝗻𝗲𝗰𝘁𝗶𝘃𝗶𝘁𝘆:

• Online

The payment is authorized immediately

Example: paying in a supermarket - the transaction is approved instantly

• Offline

The interaction happens instantly, but the actual processing happens later

Example: Tapping a card or phone at a metro gate - access is instant but the fare is charged later

𝗧𝗿𝗲𝗻𝗱𝘀 𝗮𝗻𝗱 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• NFC is not a payment method - it is an interface for initiating payments

• It standardises the interaction (the tap) but not the underlying payment. The real competition sits behind it (banks, card networks, instant payment rails, wallets)

• Because the tap looks the same everywhere, different payment types (card, wallet, A2A) can look identical - differentiation shifts to what happens after (authorization, speed, cost, control)

• As A2A replicates the card experience at POS, the tap becomes independent of the underlying rail reinforcing the shift to processing and economics

• As acceptance moves to software (phones as terminals) and offline expands reach, distribution is shifting to where NFC is embedded (e.g. smartphones, POS devices)

Opinions: Panagiotis Kriaris, Graphic source: World Bank Group

Subscribe now

4) From …Shoes to AI

A struggling shoe company makes a pivot to AI. The stock jumps by 600%+. No AI products, just the announcement.

What does this tell us about the state of the industry?

𝗧𝗵𝗲 𝘀𝘁𝗼𝗿𝘆:

Allbirds is a company that made wool sneakers - and for years, it struggled after going public.

What happened next is almost surreal.

It effectively exited its core business, selling its brand, assets, and operations for around $39mn.

Then, almost immediately, it announced a complete pivot:

• Rebranding to NewBird AI

• Raising $50mn to buy GPUs and lease AI compute capacity

• Positioning itself as an AI infrastructure / GPU-as-a-service provider

No products.

No customers.

No existing AI capability.

Just a plan.

The result?

The stock jumped 600%+ in a single day, briefly adding tens of millions in market value.

𝗜𝗻 𝘀𝗵𝗼𝗿𝘁:

The mere possibility that a company could reposition itself into the AI value chain is enough to trigger massive revaluation. Whether it can actually deliver is almost secondary, at least in the short term.

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

Allbirds is not the first company looking to cash in on the AI boom. And it will certainly not be the last.

History is full of struggling companies pivoting to the industry that is in the headlines.

• In the dot‑com era, companies just added “.com” to their name.

• In the crypto cycle, they attached themselves to blockchain to change how the market talked about them.

• More recently, we have seen pivots into electric vehicles, the metaverse, or hydrogen - often before they had any real capabilities.

𝗪𝗵𝗮𝘁 𝗶𝘀 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝘁𝗵𝗶𝘀 𝘁𝗶𝗺𝗲?

• The AI narrative is built on real, capital‑intensive infrastructure (chips, data centers, power), so even speculative pivots can appear to sit on a credible, if distant, path to revenue.

• Markets are essentially giving companies a second‑hand ticket into the AI value chain, rewarding any plausible link to AI infrastructure or compute access with a big valuation bump, even without proven products or customers.

• Because the bulk of the economic upside is already concentrated at the top of the AI stack (cloud providers, chipmakers, and core platforms), weaker incumbents are forced to reposition themselves into AI‑adjacent spaces just to regain investor interest and premium pricing.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

5) Core Skills in 2030

The future of work is here. 39% of current skill sets will be obsolete by 2030; 59% of workers will need reskilling. Understanding the direction is critical.

These are my key take-aways from the World Economic Forum’s future of jobs report:

𝗧𝗵𝗲 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• 22% of current jobs (∼ 262 million) will be affected.

• 170 million new jobs will be created and 92 million displaced, netting 78 million new roles.

𝗧𝗼𝗽 𝗗𝗿𝗶𝘃𝗲𝗿𝘀:

• Broadening digital access – 60% of employers expect it to reshape business by 2030.

• AI and tech advances – 86% cite AI as transformative; robotics (58%) and energy tech (41%) also key.

• Green transition – Climate action is driving demand for renewable energy and environmental roles.

• Demographic shifts – Aging boosts healthcare jobs; young populations fuel education demand.

𝗔𝗜’𝘀 𝗗𝗼𝘂𝗯𝗹𝗲 𝗜𝗺𝗽𝗮𝗰𝘁:

• While AI is expected to displace jobs, two-thirds of employers also plan to hire for AI-specific roles.

• Task automation will rise, but augmentation (human + machine collaboration) is seen as the more sustainable path.

𝗙𝗮𝘀𝘁𝗲𝘀𝘁 𝗚𝗿𝗼𝘄𝗶𝗻𝗴 𝗝𝗼𝗯𝘀:

• Big Data Specialists, AI/Machine Learning Specialists, FinTech Engineers, and Renewable Energy Engineers.

𝗙𝗮𝘀𝘁𝗲𝘀𝘁 𝗗𝗲𝗰𝗹𝗶𝗻𝗶𝗻𝗴 𝗝𝗼𝗯𝘀:

• Data Entry Clerks, Bank Tellers, Administrative Assistants, and Postal Service Clerks.

𝗖𝗼𝗿𝗲 𝗦𝗸𝗶𝗹𝗹𝘀 𝗼𝗳 𝘁𝗵𝗲 𝗙𝘂𝘁𝘂𝗿𝗲:

• Analytical thinking, resilience, adaptability, leadership, and AI/tech literacy.

• Curiosity and lifelong learning are rising in importance, with over 60% of employers identifying them as key skills for future roles—critical for adapting to evolving technologies and career pathways.

𝗪𝗼𝗿𝗸𝗳𝗼𝗿𝗰𝗲 𝗣𝗿𝗶𝗼𝗿𝗶𝘁𝗶𝗲𝘀:

• 85% of employers plan to upskill staff.

• Employee health and well-being is now a key talent attraction strategy.

Summary: Panagiotis Kriaris, Source: World Economic Forum

Subscribe now

The 2026 Worldpay Global Payments Report is out, and it’s a must-read for anyone who wants to understand payments. This is my analysis.

1. Global Overview:

• Wallets are the entry point to commerce (56% of e-com and 33% of POS), aggregating cards, A2A, and alternative rails.

• Cards are shifting from front-end to infrastructure, still driving 48% of POS and 31% of e-com.

• A2A wins when tied to domestic infrastructure - but fragmentation limits global scale.

• E-com is growing faster (7.5% CAGR vs 3.4% POS), concentrating future value in digital-native methods in online journeys.

2. Regional Comparison:

• Western markets remain card-heavy, while Asia is already operating on alternative rails

• A2A remains single-digit globally, but has scaled in LATAM (20%+) and MEA (15%), tied to domestic systems

• Growth is skewed to MEA (11%), LATAM (9%), APAC (8.5%) vs. Europe/NA (~6-7%), shifting global share over time

3. POS: From Terminals to Apps:

• Payment apps scale faster than the market (8% vs 3.4% CAGR) – to reach 46% of POS by 2030

• APAC has already made the shift (China 89%, India 65%, Thailand 56% POS via apps)

• QR codes remove infrastructure dependency, accelerating adoption in APAC and LATAM

• Europe is opening up, shifting from a closed card system to a competitive app layer

4. A Multipolar Landscape:

• Scale is shifting to domestic networks (UPI, PIX, Alipay), reducing reliance on global card schemes

• Cross-border is becoming a connectivity problem, solved by linking local systems (vs. expanding global ones)

• Adoption scales regionally first (APAC corridors), not globally, reinforcing fragmentation

• Europe is the exception, attempting a unified layer (Wero) instead of connecting existing systems

5. Wallets as the Control Layer:

• Wallet funding reflects local payment norms: cards in the West, A2A in markets like India, and local methods in each region

• What changes is not the rail, but controls - wallets sit on top and decide how it’s used

• As non-card rails grow, wallets become the integration layer, expanding into super apps where payments power broader ecosystems

6. BNPL as a Core Wallet Component:

• BNPL is no longer standalone but a standard feature inside wallets and checkout

• Shift from transaction monetization to lifecycle ownership, expanding into accounts, cards, and ecosystems

• The paradox: instead of replacing cards, BNPL drives installment demand back onto card rails

7. The Crypto Integration:

• Still niche in direct use (0.2% of e-com), but growing fast (16% CAGR), with scale coming via fiat-linked flows (vs. native crypto payments)

• Adoption through integration, not replacement: cards, wallets and stablecoins bridge crypto into existing rails, especially in x-border and B2B

Analysis: Panagiotis Kriaris, source: Worldpay Global Payments Report 2026 - Link to the report: https://worldpay.com/en/global-payments-report

Subscribe now

2) Challenger Banks US Exits

The US is a graveyard for challenger banks. Monzo is the latest to exit, announced last week. But why do so many challengers fail in the US?

I have looked at the biggest failures over the past years and have identified 4 patterns:

𝟭. 𝗦𝘁𝗿𝗼𝗻𝗴 𝗽𝗿𝗼𝗱𝘂𝗰𝘁, 𝗻𝗼 𝗱𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻:

Monzo and N26 had strong products but couldn’t scale distribution.
Customer acquisition in the US is expensive, and incumbents already good enough for most users. Without a cost-efficient way to acquire customers at scale, the model doesn’t work.

𝟮. 𝗟𝗶𝗺𝗶𝘁𝗲𝗱 𝗿𝗲𝘃𝗲𝗻𝘂𝗲 𝗺𝗼𝗱𝗲𝗹:

Free accounts and interchange don’t generate enough revenue per user. PerkStreet Financial, Moven and Zero Financial show that without lending or higher-margin products, the economics cannot become sustainable.

𝟯. 𝗗𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝘆 𝗼𝗻 𝘀𝗽𝗼𝗻𝘀𝗼𝗿 𝗯𝗮𝗻𝗸𝘀:

Many challengers rely on partner banks for infrastructure and balance sheet. Simple and Azlo were shut down when strategy changed, regardless of their own traction.

𝟰. 𝗡𝗶𝗰𝗵𝗲 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝘀𝗰𝗮𝗹𝗲:

Focused propositions solve real problems but often lack scale. Denizen and Daylight show that a clear use case is not enough if the market is too small to sustain the model.

There are several reasons why these patterns keep repeating themselves:

𝟭. 𝗗𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻:

There is no single national channel to acquire customers at scale, while distribution is concentrated in a few large banks controlling ~40% of deposits. Customer acquisition relies on paid channels and brand, making it costly and slow to build meaningful volume.

𝟮. 𝗥𝗲𝘃𝗲𝗻𝘂𝗲 𝗺𝗼𝗱𝗲𝗹:

Interchange alone is not sufficient, and free accounts are expected. The U.S. is a credit-driven market, so meaningful revenue comes from lending and deposit monetization - areas that require capital, risk management, and regulatory approval.

𝟯. 𝗥𝗲𝗴𝘂𝗹𝗮𝘁𝗶𝗼𝗻:

Overlapping federal and state regulation makes licensing slow and expensive. Most challengers operate through sponsor banks that hold the license, deposits, and customer funds, leaving them dependent on partners for core capabilities and strategic decisions.

𝟰. 𝗛𝗶𝗴𝗵 𝘀𝗰𝗮𝗹𝗲 𝘁𝗵𝗿𝗲𝘀𝗵𝗼𝗹𝗱𝘀:

Fixed costs for compliance, operations, and support are significant from the start, while fragmented payment rails (cards, ACH, Zelle, FedNow) add complexity. Without large volumes the model cannot absorb these costs.

𝗦𝗼, 𝘄𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝗶𝘁 𝘁𝗮𝗸𝗲 𝘁𝗼 𝘄𝗶𝗻?

• Distribution needs to be built in early and not rely on paid acquisition.
• Revenue needs to go beyond payments (lending, deposits).
• Control over key parts of the stack, especially the balance sheet.
• Fast scaling - given the high cost base from day one.
• Embedding the product into everyday financial activity.

Who can make it work?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) Global Stablecoin Regulation

US$33 trillion moved through stablecoins in 2025 - more than Visa and Mastercard combined. Regulators are stepping in - but their approaches are very different.

Still, 𝘁𝗵𝗿𝗲𝗲 𝗰𝗼𝗺𝗺𝗼𝗻 𝘁𝗵𝗲𝗺𝗲𝘀 are emerging:

• Reserves: stablecoins must be fully backed 1:1 by safe, liquid assets - typically cash or short-term government bonds - to preserve value and ensure redemption.

• Redemption rights: users must be able to cash out at face value, within timelines set by law - ranging from same-day (UAE, Hong Kong) to five days (Singapore).

• Independent custody: backing assets must be held separately from the issuer’s own funds, often by regulated custodians or in trust, to protect users in case of failure.

These shared principles reflect regulatory alignment on the minimum requirements for trust and stability in issuing and using stablecoins.

𝗕𝘂𝘁 𝗵𝗼𝘄 𝘁𝗵𝗲𝘆’𝗿𝗲 𝗶𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝗲𝗱 𝘃𝗮𝗿𝗶𝗲𝘀 𝘄𝗶𝗱𝗲𝗹𝘆:

• Who can issue: some jurisdictions restrict this to banks (Japan, South Korea), while others permit non-bank fintechs (EU. Hong Kong, Singapore).

• Reserve rules: the US allows only cash and Treasuries; others like Japan and the UK permit a broader mix of safe assets.

• Redemption timelines: these differ significantly - affecting liquidity and user expectations.

• Cross-border limits: Some regimes block foreign-issued stablecoins unless they meet local regulatory standards (e.g. EU, UAE).

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• The US push is accelerating adoption - but puts pressure on non-US issuers to either comply with US rules or exit the market.

• Asia’s bank-led models favour control and stability but may limit openness and cross-border scale.

• UK–EU regulatory alignment will determine whether stablecoins can move freely between key markets - or remain siloed.

𝗪𝗵𝗮𝘁’𝘀 𝗻𝗲𝘅𝘁:

• Stablecoin regulation is unfolding much like the early days of card networks - built jurisdiction by jurisdiction, with each market defining its own rules on issuance, custody, reserves, and redemption.

• Alignment may come, but not soon. Meanwhile, adoption is accelerating. Trillions are already flowing through stablecoins, and regulators are shifting from drafting rules to enforcing them.

• For issuers and infrastructure providers, waiting for harmonisation is a risk. Competing in this space means navigating a complex patchwork of rules, or losing access to key markets.

Opinions: Panagiotis Kriaris, Graphic source and data points: EY

Subscribe now

4) OpenAI vs. Anthropic

It is the embodiment of the AI race. OpenAI vs. Anthropic. But what looks like two similar models is in fact two fundamentally different strategic plays.

𝗢𝗽𝗲𝗻𝗔𝗜:

• OpenAI built its strategy around the interface. ChatGPT is not just a product, it is the distribution layer where consumer and enterprise meet, and where access, usage, and pricing are controlled.

• This puts OpenAI in control of demand. It does not depend on partners to drive usage and can decide how AI is used and monetized.

• The trade-off: building and maintaining that demand is expensive, and revenue grows with user adoption (vs. enterprise clients).

𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗶𝗰:

• Anthropic made the opposite bet and prioritized enterprise distribution over interface ownership, embedding its models into platforms that already have scale (Amazon, Google).

• This allows immediate monetization. Usage comes from existing enterprise workflows, often at scale, without the need to build a consumer product first.

• The result is faster revenue growth. Instead of competing for attention, Anthropic sits where budgets already exist and where usage converts directly into spend.

𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝗰𝗼𝗺𝗽𝗮𝗿𝗶𝘀𝗼𝗻:

• Anthropic is now at $30bn annualised revenue, versus $24bn for OpenAI. A year ago, it was $1bn vs $6bn!

• The reversal is driven by monetization. Anthropic generates the majority of its revenue from enterprise (80%), while OpenAI is at 40%. One converts usage into revenue early, the other builds usage first and monetizes over time.

• This shows in how revenue is generated. Anthropic has 1,000+ customers spending $1m+ annually, driven by deployment in coding, operations, and internal workflows. OpenAI operates at much larger scale (~900mn weekly users), but a significant share of that usage is free or low-priced.

• Both sell their AI through cloud platforms (Microsoft for OpenAI, Amazon and Google for Anthropic), which handle infrastructure and distribution and take a share of the revenue.

• OpenAI reports only the portion it keeps after paying Microsoft, while Anthropic reports the full amount customers pay before sharing with its partners - so Anthropic’s revenue looks higher, even though a meaningful part of it goes to Amazon and Google.

But despite the accounting difference 𝘁𝗵𝗲 𝗯𝗼𝘁𝘁𝗼𝗺 𝗹𝗶𝗻𝗲 𝗶𝘀 𝗰𝗹𝗲𝗮𝗿:

A not long ago relatively unknown company, founded years after OpenAI by former OpenAI employees, has in a short time caught up and, on some metrics, moved ahead of the company that defined consumer AI.

How? Enterprise-first.

Now OpenAI is moving in the same direction. And this will be the name of the game going forward.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

5) The evolution of Software

Software is at a crossroads. Having driven disruption across industries, it is now facing disruption of its own business model.

But within that shift, one aspect is still being missed.

Software evolution unfolded across four phases.

𝟭. 𝗢𝗻-𝗽𝗿𝗲𝗺𝗶𝘀𝗲

• Software is owned and operated within company infrastructure, with control sitting in IT

• Systems are built for stability, with long lifecycles and deep customization

• Adoption and change are slow, centrally governed, and tightly managed

𝟮. 𝗦𝗮𝗮𝗦

• Software becomes cloud-delivered and subscription-based, lowering the barrier to adoption

• Decision-making shifts from IT to business teams

• The number of tools grows rapidly, often without clear visibility

𝟯. 𝗖𝗹𝗼𝘂𝗱 & 𝗔𝗣𝗜𝘀

• Software evolves into modular, composable systems built on infrastructure and APIs

• Products are assembled (not bought), with engineering defining architecture and scale

• Software becomes tightly integrated into how products are built and delivered

𝟰. 𝗔𝗜

• Software is increasingly built around models, data, and workflows

• Systems are used to generate outputs, support decisions, and automate

• Software becomes embedded across products, operations, and processes

𝗘𝗮𝗰𝗵 𝗽𝗵𝗮𝘀𝗲 didn’t just follow technology or business needs - it introduced a new monetization model.

That shift reshaped spend - from fixed and predictable to distributed and usage-driven.

And there is a clear one-to-one correspondence between 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝘀𝗽𝗲𝗻𝗱:

• On-premise: Spend was upfront, predictable, and centrally controlled – tied to long planning cycles

• SaaS: Spend shifted to recurring and decentralized - enabling faster adoption, but reducing visibility and teams coordination

• Cloud & APIs: Spend became usage-based and directly linked to how systems were built and run.

• AI: Spend is becoming activity-driven - shaped by how much work software executes, and increasingly dependent on product, engineering, and operational choices

𝗔𝗻𝗱 𝗵𝗲𝗿𝗲 𝗶𝘀 𝘁𝗵𝗲 𝗽𝗮𝗿𝗮𝗱𝗼𝘅:

Most of the focus today is on changing business models, new revenue opportunities, and the role of AI, but much less attention is paid to the operational side.

Yet this is exactly where a large share of company spend is now determined - by how these systems are actually run.

A new breed of companies is solving for this. Companies like Spendbase bring visibility and control to how software and cloud spend are managed across the organization:

• Consolidate software and cloud spend into a single view, across vendors, teams, and environments

• Identify unused licenses, overlapping tools, and underutilized resources

• Link usage to cost, showing what is driving spend across SaaS, cloud, and APIs

The shift to how software is run is the least discussed, yet the one with the most impact.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

1) Agentic Commerce Protocols

Agentic commerce is having its HTTP moment. Three-letter protocols are competing to define the rules for the next big thing - just like HTTP did for the internet.

A protocol is a set of rules that defines how systems interact with each other:

• HTTP defines how a browser loads a webpage - the foundation of the internet

• Agentic protocols define how AI agentσ turn requests into actions.

And just like the internet, it won’t be one protocol - it will be a stack like HTTP, TCP/IP, and DNS together.

In practice, this means defining:

• Discovery: How agents find options: like DNS find websites

• Intent: How agents express choices: like HTTP structures requests

• Execution: How agents trigger actions: like TCP ensures reliable delivery

• Settlement: How results complete: like IP routes data end-to-end

Although there are more, 4 main protocols have emerged:

𝗔𝗖𝗣 (𝗔𝗴𝗲𝗻𝘁 𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗣𝗿𝗼𝘁𝗼𝗰𝗼𝗹)

• Defines how different agents communicate with each other

• Allows one agent to pass context, ask another for input, or split a task, so the outcome is coordinated

• Similar to teams agreeing on what needs to be done and who does what

𝗨𝗖𝗣 (𝗨𝗻𝗶𝘃𝗲𝗿𝘀𝗮𝗹 𝗖𝗼𝗺𝗺𝗲𝗿𝗰𝗲 𝗣𝗿𝗼𝘁𝗼𝗰𝗼𝗹)

• Defines how an agent can look across multiple providers

• Allows the agent to compare options, evaluate, and choose the best option

• Similar to checking different websites or shops before deciding what to buy

𝗠𝗖𝗣 (𝗠𝗼𝗱𝗲𝗹 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗣𝗿𝗼𝘁𝗼𝗰𝗼𝗹)

• Defines how businesses expose what they can do so an agent can actually use them

• Allows an agent to trigger real actions such as booking a ticket without custom integrations each time

• Similar to filling in a form and completing the action

𝗠𝗣𝗣 (𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗣𝗮𝘆𝗺𝗲𝗻𝘁 𝗣𝗿𝗼𝘁𝗼𝗰𝗼𝗹)

• Defines how an agent completes the transaction

• Allows payments to be initiated, routed, and settled automatically without human intervention

• Similar to entering payment details and completing the checkout

And this is why we see different players behind each protocol - they’re not solving the same problem, they are trying to control different parts of the flow:

• AI platforms (OpenAI, Anthropic) are pushing protocols to move beyond chat, because without controlling coordination and execution they remain interfaces

• Google is extending what it already does, discovery and comparison, into transactions, so that it remains in the flow and not being bypassed by agents

• Payments players (Stripe) are embedding themselves into how transactions are completed, because if agents remove checkout, payments risk becoming invisible and commoditized.

𝗪𝗵𝗮𝘁’𝘀 𝗺𝗶𝘀𝘀𝗶𝗻𝗴?

Players that connect these layers end-to-end.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) The Social-Finance Flywheel

This is one of the most strategic moves lately. TikTok is moving to become a financial player in Brazil.

𝗪𝗵𝗮𝘁 𝗧𝗶𝗸𝗧𝗼𝗸 𝗵𝗮𝘀 𝗱𝗼𝗻𝗲:

TikTok has reportedly applied for two licenses with Brazil’s central bank:

• An electronic money issuer: to offer prepaid ​accounts for users to hold balances, receive funds and make payments within ​its app

• A direct credit company: to lend its own capital or act as a platform connecting borrowers and lenders

𝗧𝗵𝗲 𝗠𝗼𝗱𝗲𝗹:

This is not a new model. Douyin (TikTok’s sister app in China) already does this successfully:

• Credit is embedded directly into the consumption journey - financing appears at the moment of intent (shopping, live streams, creator recommendations), turning engagement into transactions

• Distribution is algorithm-driven with credit offers triggered by behavior, context, and conversion probability, using platform data alongside traditional signals for underwriting

• Credit increases conversion, basket size, and merchant demand, making it a direct driver of sales

𝗪𝗵𝘆 𝗕𝗿𝗮𝘇𝗶𝗹:

• A real-time payment system (PIX) is used at national scale, making instant, low-cost transactions part of everyday life and creating a natural entry point to layer credit on top

• Large underserved credit demand, especially among consumers and SMEs with limited access to traditional financing

• High usage of social media and strong engagement with creators, creating a natural link between content, commerce, and financing

• Regulatory framework that supports fintechs, with open banking and licensing models that allow new players to enter financial services

• High interest rates and lending margins, making the economics of credit attractive for new entrants

Outside China, TikTok has not been able to move beyond limited or partnership-based setups, which kept it at the distribution layer

In Brazil, it is aiming to own the financial layer by applying for both payments and lending licenses, moving closer to underwriting, pricing, and embedding credit directly into the platform

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• Social platforms are moving from just driving traffic to capturing the economics behind it, turning attention into transactions and transactions into financial revenue

• Credit is introduced at the moment of purchase to influence the outcome, increasing approval, conversion, and basket size rather than being set up separately beforehand

• The key advantage is not just data, but owning the full flow from discovery to checkout, where demand is created and captured in the same environment

• If this works, it will create a repeatable model (outside China) where content, commerce, payments, and credit reinforce each other in one loop, raising the bar for both banks and fintechs

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) Digital Banking Economics

The old digital banking economics playbook no longer works.

To understand what does, we need to look at what has fundamentally changed. Three core shifts stand out:

𝗦𝗵𝗶𝗳𝘁 𝟭:

• Digital banks are shifting from prioritising user growth toward generating more value per customer.

• This means expanding into credit, wealth, and SME services, where revenue is linked to ongoing financial activity, not single transactions.

• The shift is from number of users to revenue depth per user, requiring tighter product integration and consistent engagement.

𝗦𝗵𝗶𝗳𝘁 𝟮:

• The front-end app is no longer enough - banks are building core capabilities to participate in lending, deposits, and risk-taking.

• This introduces balance sheet economics, where revenue comes from spreads and capital deployment (vs. transaction fees).

• Moving into this layer requires licenses, underwriting, and operational discipline.

𝗦𝗵𝗶𝗳𝘁 𝟯:

• Digital banks are shifting from isolated products to integrated ecosystems.

• The goal is to integrate into daily workflows, especially for SMEs, so the bank becomes part of how money is managed (vs. stored).

• This creates repeat usage and stronger retention, enabling monetisation across multiple touchpoints instead of a single product.

𝗧𝗵𝗲 𝗻𝗲𝘄 𝗺𝗼𝗻𝗲𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹:

• The big change is where revenue is generated. Instead of transactions, revenue moves to the financial position held - balances, credit exposure, and how often it is used.

• This makes revenue structurally different. Transaction-based income depends on volume, whereas position-based income builds over time and is harder to compete away because it sits inside the customer’s financial setup.

• The consequence is that two players can look similar on the surface, but generate very different economics depending on whether they control flows or positions.

• This is also why the shift is difficult to execute. Moving into these revenue pools requires owning risk, managing capital, and operating with discipline, which most front-end led models cannot support.

𝗘𝘅𝗮𝗺𝗽𝗹𝗲𝘀:

• Revolut uses payments and FX to drive activity, but revenue comes from adding credit, subscriptions, and wealth on top, so income continues beyond the transaction. Wise remains tied to money movement, which limits revenue per customer.

• Qonto is embedded in SME workflows like invoicing and expenses, so revenue grows with how the business operates. Retail neobanks stay closer to cards and payments, keeping income linked to activity.

• OakNorth generates revenue by deploying capital with disciplined underwriting, leading to high revenue per client. Most front-end fintechs cannot move into this layer because they do not have the risk or balance sheet capabilities.

Opinions: Panagiotis Kriaris, Graphic source: C-Innovation

Subscribe now

4) The Card Issuing Gap

Card issuing has a real gap. A legacy operating model built for banks - never designed for the platform economy.

𝗧𝗵𝗲 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗶𝘀𝘀𝘂𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹:

• Done in-house by banks on proprietary systems and owned processing infrastructure

• Rigid, end-to-end model with little flexibility or customization

• High costs driven by heavy upfront investment in technology and infrastructure

𝗧𝗼𝗱𝗮𝘆, 𝗺𝗮𝗿𝗸𝗲𝘁 𝗻𝗲𝗲𝗱𝘀 𝗮𝗿𝗲 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹𝗹𝘆 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁:

• Companies need to embed cards as their own product, without managing the underlying complexity

• Card programs must launch in weeks rather than months or years

• Issuing must be modular and API-driven, enabling control over UX, features, and flows.

• Cost structures must be variable and scalable, without heavy upfront investment.

This is exactly what I call the 𝘁𝗵𝗲 𝗰𝗮𝗿𝗱 𝗶𝘀𝘀𝘂𝗶𝗻𝗴 𝗴𝗮𝗽:

The mismatch between what digital businesses need and what legacy issuing models can deliver.

A new breed of providers is rethinking issuing around modern digital businesses, platforms, marketplaces, and embedded use cases.

𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀:

𝟭. Cloud-based, modular architecture, allowing issuing capabilities to evolve without rebuilding the stack

𝟮. Embedded model spanning both cards and accounts (CaaS / BaaS)

𝟯. API-driven management of cards and accounts, enabling direct integration and automation

𝟰. High configurability: real-time controls, spending rules, branded card/account experiences

𝟱. Unbundled infrastructure, with specialised parties handling card production, processing, and backend operations

𝟲. EMI-licensed setup, supporting electronic money accounts alongside card programs

𝟳. Flexible, usage-based pricing, avoiding heavy upfront technology and infrastructure costs

𝟴. Built-in modern security, including tokenisation, encryption, and digital wallet support (Apple Pay, Google Pay)

𝟵. End-to-end issuing coverage, from onboarding and personalisation to authentication, risk, and fraud management

Implications:

• Issuing is moving from a banking function to a product capability, embedded directly into platforms and user journeys where payments are part of the experience

• Control shifts from infrastructure ownership to API-level control, allowing companies to define limits, authorization logic, and user experience in real time

• Differentiation moves from access to cards to how issuing is designed and monetized - through spend controls, embedded credit, incentives, and integration into product flows

• Value comes from combining issuing with accounts, data, and workflows, enabling use cases like automated expense management, marketplace pay-outs, SaaS spend controls, or financing directly within the transaction flow

Opinions and graphics: Panagiotis Kriaris

Subscribe now

1) GenAI Reference Architecture

GenAI is easy to start but hard to scale. Too many companies are stuck in endless pilots. Here’s what it takes to build Gen AI capability.

McKinsey has recently published their findings from working with 150+ companies on their GenAI programs over two years.

Two hurdles stand out:

𝟭. 𝗙𝗮𝗶𝗹𝘂𝗿𝗲 𝘁𝗼 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗲: Teams waste time on duplicate experiments, wait on compliance processes, and solve problems that don’t matter. 30% - 50% of innovation time is spent trying to meet compliance - not building.

𝟮. 𝗙𝗮𝗶𝗹𝘂𝗿𝗲 𝘁𝗼 𝘀𝗰𝗮𝗹𝗲: Even when a prototype works, most companies can’t get it into production. Risk, security, and cost barriers overwhelm teams, leading to stalled or cancelled deployments.

According to McKinsey the most successful GenAI platforms contains three core components:

𝟭. 𝗔 𝘀𝗲𝗹𝗳-𝘀𝗲𝗿𝘃𝗶𝗰𝗲 𝗽𝗼𝗿𝘁𝗮𝗹:

To support both innovation and scale, companies need a secure, centralized portal that gives teams easy access to pre-approved gen AI tools, services, and documentation. It should enable developers to quickly build with reusable patterns, while also offering governance features like observability, cost controls, and access management. The best portals promote contribution and reuse across the organization, reducing friction and accelerating development at scale.

𝟮.𝗔𝗻 𝗼𝗽𝗲𝗻 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝘁𝗼 𝗿𝗲𝘂𝘀𝗲 𝗚𝗲𝗻𝗔𝗜 𝘀𝗲𝗿𝘃𝗶𝗰𝗲𝘀:

Scaling GenAI requires modular, open architecture that enables teams to reuse services, application patterns, and data products across use cases. Leading companies build libraries of common components (like RAG, embeddings, or chat workflows) and focus on integration via APIs - not vendor lock-in. Infrastructure and policy as code ensure changes can propagate quickly and securely across the platform, reducing cost and accelerating deployment.

𝟯. 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱, 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗹𝗲 𝗔𝗜 𝗴𝘂𝗮𝗿𝗱𝗿𝗮𝗶𝗹𝘀:

To scale safely, GenAI platforms must embed automated governance that enforces compliance, manages risk, and tracks costs. This includes microservices that audit prompts, detect policy violations (like sharing sensitive personal data or generating inaccurate responses), and attribute usage to specific teams. A centralized AI gateway enforces access controls, logs interactions, and routes traffic through security filters - allowing flexibility where needed. These guardrails accelerate approval processes, reduce setup time, and let teams focus on building value - not managing risk manually.

Summary: Panagiotis Kriaris, Source: McKinsey & Company

Subscribe now

2) Revolut 2025 vs. 2024

The Revolut 2025 Annual Report is out. Here’ s what’s behind the numbers.

𝗡𝘂𝗺𝗯𝗲𝗿𝘀’ 𝗔𝗻𝗮𝗹𝘆𝘀𝗶𝘀:

• Proven profitability at scale:

Revolut’s profit before tax grew from £438M in 2023 to £1.1B in 2024 to £1.7B in 2025 - proof that a pure-digital bank can be a capital-generating platform at scale.

• Expanding margins:

Revenue grew 46% from £3.1B to £4.5B - but profit before tax grew 57% over the same period, indicating that unit economics have tightened year-on-year and operating leverage is now a clear structural feature.

• Mix toward higher‑value segments:

Card, FX and Revolut Business now make up a meaningful share of revenue, with the SME/B2B layer growing faster than consumer. That means the business is deliberately rebalancing toward higher‑margin, stickier flows.

• The real moat is not the app:

The fastest growth in Revolut Business and higher‑margin flows versus 2024 shows the company is embedding into SME and B2B workflows, effectively using the app as a gateway to relationship‑based banking (and not just serving as a consumer wallet).

• Economics shifting from acquisition to monetization:

With customer growth moderating versus 2024 but revenue and profit accelerating in 2025, Revolut is clearly monetizing deeper into the existing base - the 2024 base is now the core profit engine.

• Strategic positioning:

The 2025 numbers reposition Revolut as a high‑growth, capital‑light infrastructure play that is no longer just challenging incumbents on user experience, but actively pressuring their spreads and cost structures via a more efficient, software‑driven model.

𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• Revolut’s shift from user‑growth to profit‑and‑mix‑driven expansion sets a new template for how investors will judge other neobanks: scale plus margin profile and not just how big their user base looks on the surface.

• The fact that Revolut is monetizing deeper into its existing base, not just adding users, signals that the era of growth‑only neobanks is over. The next phase is about embedding, cross‑selling and relationship depth.

• The strong shift toward SME/B2B flows suggests that the real play in digital‑banking stacks lies in business‑embedded finance (vs. competing head‑on with incumbents on mass‑market retail accounts).

• For incumbents, Revolut’s capital‑light model is less an existential threat today than a long‑term margin‑compression force: as more flows move onto efficient digital rails, spreads and cost bases will erode gradually (but not overnight).

• For other fintechs, the key takeaway is that standalone wallets or narrow payments plays are no longer enough. The winners will be those that can layer full‑stack, B2B‑oriented rails on top of a large, sticky user base.

𝗪𝗵𝗮𝘁 𝗶𝘀 𝘆𝗼𝘂𝗿 𝗯𝗶𝗴𝗴𝗲𝘀𝘁 𝘁𝗮𝗸𝗲-𝗮𝘄𝗮𝘆?

Opinions and graphics: Panagiotis Kriaris, Numbers’ source: Revolut Annual Reports

Subscribe now

3) The Evolution of Identity

Identity is the most misunderstood layer in financial services.

Now it is becoming core infrastructure.

Historically, financial services scaled through standardization and shared infrastructure. Identity has followed a different path:

𝟭. 𝗠𝗮𝗻𝘂𝗮𝗹 𝗶𝗱𝗲𝗻𝘁𝗶𝘁𝘆

Identity was established directly between the institution and the customer, typically in person.

This ensured trust and control.

But it limited scale, tying growth to physical presence.

𝟮. 𝗗𝗶𝗴𝗶𝘁𝗮𝗹 𝗶𝗱𝗲𝗻𝘁𝗶𝘁𝘆

Identity moved online, allowing institutions to onboard customers remotely and expand distribution.

This solved physical constraints.

But each institution built its own processes, leading to repeated verification and duplication.

𝟯. 𝗩𝗲𝗿𝗶𝗳𝗶𝗲𝗱 𝗶𝗱𝗲𝗻𝘁𝗶𝘁𝘆

Specialized providers and national identity schemes improved the quality and speed of verification.

This made onboarding more reliable and easier to scale.

But instead of convergence, it created a fragmented set of identity sources across markets.

𝟰. 𝗜𝗱𝗲𝗻𝘁𝗶𝘁𝘆 𝗮𝘀 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲

The number of identity systems increased, improving coverage for institutions operating globally.

This gave access to more trusted identity sources.

But it shifted the problem from verification to integration.

𝗔𝗻𝗱 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗲𝘅𝗮𝗰𝘁𝗹𝘆 𝘁𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺:

As financial services become more cross-border, embedded, and real-time, identity is no longer a one-off step.

It sits across onboarding, access, fraud, and compliance - but remains fragmented.

There are now 150+ government-backed digital identity schemes worldwide, each built for a specific market, with no common standard.

𝗪𝗵𝗶𝗰𝗵 𝗺𝗲𝗮𝗻𝘀:

• Every new market comes with a new identity integration

• The same user goes through KYC again and again, depending on the provider and country

• Risk and fraud decisions change by market because the underlying identity data is different

• The product experience breaks across countries

𝗧𝗵𝗲 𝗻𝗲𝗲𝗱:

It’s obvious that there is a clear need for orchestration.

Much like payments moved from local, disconnected systems to global networks like Visa and Mastercard - replacing many individual connections with a single shared layer - identity is now going through the same shift.

Hopae is a good example of a company building identity as the next infrastructure layer - a digital identity layer for the internet:

• Access to multiple identity schemes through a single integration

• Reusable identity across providers and countries, instead of repeated verification

• Consistent identity signals across onboarding, authentication, and risk

• Identity as an access layer, not a system to integrate

This is one of the biggest shifts in finance in decades: identity is becoming the protocol layer of global finance - not moving money but deciding who can.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

4) Agent-to-Agent Communication

The biggest disruption of agents in finance is not about AI models. It’s about communication. Yet very few are preparing for it.

Today, financial decisions take place in apps and websites, where customers must search and compare offerings.

In an agentic economy, financial institutions will increasingly interact with AI agents rather than customers directly.

𝗪𝗶𝘁𝗵 𝗽𝗿𝗼𝗳𝗼𝘂𝗻𝗱 𝗰𝗼𝗻𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲𝘀:

• Products must become machine-readable. If agents cannot access the information they need, they will simply leave them out.

• Distribution will shift from interfaces to protocols as products will need to be accessible through APIs and agent frameworks.

• This will be a 24/7 always-on play. Agents will continuously monitor, compare, and act on behalf of customers.

• Institutions will need clear rules for what their agents can disclose, negotiate, and execute.

• Agent verification will be critical. Providers need to know who the agent represents and what actions it is authorized to take.

• AI agents will become a new layer between providers and customers.

𝗕𝘂𝘁 𝘁𝗵𝗶𝘀 𝘄𝗶𝗹𝗹 𝗻𝗼𝘁 𝗯𝗲 𝗮 𝗼𝗻𝗲-𝘀𝗶𝗱𝗲𝗱 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝗶𝗼𝗻.

Customers will have agents acting on their behalf, but providers will increasingly deploy their own agents to respond to requests, provide information, and execute transactions.

In other words, this will become an agent-to-agent communication on both sides.

In fact, this two-way communication is already being deployed. Projects such as OpenClaw are building marketplaces (check Frederic Godin’s article for Sinch) where companies publish services that AI assistants can discover and use automatically.

𝗧𝗵𝗲 𝗸𝗲𝘆 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗶𝘀:

If this is the beginning of a completely new reality, how will these agents actually communicate?

The answer lies in communication infrastructure that allows agents to exchange messages, verify identities, request confirmations, and trigger actions across different providers.

But building and operating this type of infrastructure requires specialized providers such as Sinch. In practice, this means:

• Enabling requests, confirmations, and updates triggered by agents to reach users and counterparties across messaging and voice networks.

• Handling identity and verification, ensuring that when an agent initiates an action, providers can verify who it represents and whether the request is authorized.

• Connecting automated actions to real-world communication channels, such as messaging, calls, and authentication flows.

• Integrating with emerging agent standards such as the Model Context Protocol (MCP), which allows AI agents to discover and trigger external capabilities automatically.

This is not an operational shift but a strategic one.

𝗪𝗶𝗹𝗹 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 𝗶𝗻𝗰𝗿𝗲𝗮𝘀𝗶𝗻𝗴𝗹𝘆 𝗳𝗼𝗰𝘂𝘀 𝗼𝗻 𝗶𝗻𝗳𝗹𝘂𝗲𝗻𝗰𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁𝘀 𝗿𝗮𝘁𝗵𝗲𝗿 𝘁𝗵𝗮𝗻 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝘀?

Opinions: Panagiotis Kriaris, Graphic sources: BCG, Panagiotis Kriaris

Subscribe now

5) Webinar: Banking on APIs

Imagine having a car but never using the highway. The same analogy applies to banks not leveraging the rails of the modern economy: APIs. Let’s unpack how it actually works.

There is still significant untapped value in how banks manage and use APIs - and there is no better way to understand it than through concrete use cases.

Credit is a clear example.

Traditional credit decisions rely on limited, static data, leading to a narrow and often outdated view of the customer - and, in turn, slower and more conservative outcomes.

Modern banking uses structured API access to real transaction data to analyse income, spending, and affordability in real time - enabling faster, more precise, tailored credit decisions.

Here’s how it works in a structured, API-driven environment:

𝟭. 𝗧𝗵𝗲 𝗿𝗼𝗹𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗔𝗣𝗜 𝗺𝗮𝗿𝗸𝗲𝘁𝗽𝗹𝗮𝗰𝗲 (𝗲.𝗴. 𝗗𝗶𝗴𝗶𝘁𝗮𝗹𝗔𝗣𝗜)

• The API Gateway routes requests → All transaction data calls pass through a controlled entry point.

• The API Catalog makes APIs usable → Teams identify and use account and transaction APIs directly.

• Auth & Consent handles access → Only permitted data is accessed, with every request traceable.

• The Sandbox enables testing → Credit flows are validated with simulated data before going live.

• Governance applies rules → Access follows defined policies and regulatory requirements as usage grows.

• Analytics provides visibility → The bank tracks API usage and performance across credit applications.

𝟮. 𝗪𝗵𝗮𝘁 𝘁𝗵𝗲 𝗯𝗮𝗻𝗸 𝗱𝗼𝗲𝘀

• Identifies required data → Selects account and transaction APIs to complement bureau data.

• Validates the setup → Reviews how transaction data (income, expenses, cash flow, etc) is structured and tests the full flow before going live.

• Activates and integrates APIs → Subscribes to the required APIs and embeds them into the decisioning process.

• Combines bureau and transaction data → Uses bureau scores as a baseline and enriches them with actual financial behaviour.

• Runs real-time decisioning → Assesses income stability, spending patterns, and affordability during the application.

𝟯. 𝗧𝗵𝗲 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲

• Starts a credit application → Through a digital channel, no manual steps.

• Gives consent → Approves access to account and transaction data.

• Data is retrieved in real time → 12–24 months of transaction data is pulled alongside the bureau score.

• Financial behaviour is analysed → Income, spending, and affordability are assessed.

• A decision is made instantly → Bureau data and behavioural insights are combined.

• Receives a personalised offer → Immediate response with tailored limit, rate, and terms.

Does that sound interesting?

Together with Digitalapi.ai (Srinivasan Shanmuganathan), 𝘄𝗲 𝗮𝗿𝗲 𝗵𝗼𝘀𝘁𝗶𝗻𝗴 𝗮 𝘄𝗲𝗯𝗶𝗻𝗮𝗿 to unpack use cases like this and explain how banks can benefit from API management.

𝗪𝗵𝗲𝗻: April 1st, 14.00 CET

𝗛𝗼𝘄 𝘁𝗼 𝗷𝗼𝗶𝗻: https://na2.hubs.ly/H04rgnT0

Subscribe now

1) Card Schemes vs. Digital Wallets

Digital wallets have become the biggest threat to the card schemes. Visa and Mastercard are now fighting back.

Online card payments were built around manual card entry at checkout.

𝗧𝗵𝗮𝘁 𝗺𝗼𝗱𝗲𝗹 𝗶𝘀 𝗻𝗼𝘄 𝗯𝗿𝗼𝗸𝗲𝗻.

Commerce interfaces have shifted to wallets, platform checkouts, in-app payments, and increasingly automated purchasing.

Wallets succeeded because they removed manual card entry and replaced it with stored credentials and biometrics.

As a result, consumers interact with the wallet - not the card brand behind it.

If Visa and Mastercard remain tied to manual card entry, they risk becoming invisible infrastructure behind wallets and platforms.

𝗦𝗼, 𝗵𝗼𝘄 𝗱𝗼 𝘁𝗵𝗲𝘆 𝗿𝗲𝘀𝗽𝗼𝗻𝗱?

They are rebuilding card checkout around wallet-level convenience.

Using 2 things: Click to Pay and Passkeys:

• 𝗖𝗹𝗶𝗰𝗸 𝘁𝗼 𝗣𝗮𝘆 is a universal card checkout button. It lets consumers save their card once and pay across many websites without entering card details again.

• 𝗣𝗮𝘀𝘀𝗸𝗲𝘆𝘀 are a new way to confirm payments using the phone or laptop people already use - with Face ID, fingerprint, or device unlock, instead of passwords or SMS codes.

Together, they remove the two biggest frictions of online card payments: typing card numbers and dealing with authentication codes.

• Card credentials are already stored

• Authentication is instant via biometrics

• Checkout becomes almost one-tap

𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝗼𝗯𝗷𝗲𝗰𝘁𝗶𝘃𝗲?

Control of the checkout interface.

By pushing Click to Pay + passkeys, Visa and Mastercard are trying to make card checkout as seamless as a wallet so merchants and consumers don’t need a separate wallet.

Longer term, they want this to become the default web checkout standard. If it works, it will give the networks control over identity, authentication, and credentials.

𝗕𝘂𝘁 𝘁𝗵𝗲𝗿𝗲 𝗶𝘀 𝗮𝗻 𝗮𝗱𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗳𝗮𝗰𝘁𝗼𝗿 that is becoming critical: agentic commerce.

• In a world where AI agents buy things, the checkout cannot depend on manual typing, OTP codes and passwords

• Agents need machine-readable payment credentials and authentication flows

• Click to Pay + passkeys give the networks tokenized card credentials, strong cryptographic authentication, standardized APIs

In other words, they create a programmable payment identity.

But this is not just about improving checkout UX:

• Visa and Mastercard are preparing cards for the next interface of commerce.

• Click to Pay and passkeys are their attempt to ensure cards remain at the center of payments - not hidden behind someone else’s wallet.

𝗪𝗶𝗹𝗹 / 𝗖𝗮𝗻 𝘁𝗵𝗲𝘆 𝘀𝘂𝗰𝗰𝗲𝗲𝗱?

Opinions: Panagiotis Kriaris, Graphic sources: Visa, Mastercard, Panagiotis Kriaris

Subscribe now

2) UniCredit vs. Commerzbank

Unicredit has just announced a public takeover offer for Commerzbank.

This is the biggest debate in European banking.

𝗧𝗵𝗲 𝗯𝗮𝗰𝗸𝗴𝗿𝗼𝘂𝗻𝗱:

— Over the past years, UniCredit has steadily built up its stake in Commerzbank.

— By mid-2025, it crossed the 20% mark. The latest public offer could push its stake beyond the 30% German takeover threshold.

— Commerzbank’s management has so far resisted talks of a takeover.

— The German state still owns about 12.7%, giving Berlin effective veto power.

𝗧𝗵𝗲 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

— UniCredit’s earnings power comes from years of restructuring, cost control, and a simplified business model - the result of branch cuts, tech upgrades, and a shift towards profitability.

— Despite recent profits, Commerzbank still struggles with weaker efficiency, lower returns, and reliance on the slow-growing German market - leaving it vulnerable vs. UniCredit’s more diversified model.

— UniCredit runs a business ~50% larger with less than twice the staff - pointing to higher productivity per employee. This efficiency gap underpins the main source of potential merger synergies: overlapping branches, systems, and back-office functions.

— Commerzbank’s higher P/E reflects investor perception: Germany’s backing makes its balance sheet look safer.

— Commerzbank’s lower NPE ratio reflects its concentration in Germany, where loan defaults are low - but that same focus leaves it less diversified as the German economy slows down.

— Both banks hold capital well above requirements - the debate is how much can be returned to shareholders.

𝗪𝗵𝗮𝘁 𝗮 𝗺𝗲𝗿𝗴𝗲𝗿 𝘄𝗼𝘂𝗹𝗱 𝗺𝗲𝗮𝗻:

— Creates one of Europe’s top-5 banks with ~€1.5 trillion in assets.

— UniCredit’s higher efficiency could lift Commerzbank’s returns.

— Overlapping German operations would likely mean branch closures and job reductions - highly sensitive politically, as Berlin fears losing a national champion and weakening SME financing.

𝗗𝗼𝗲𝘀 𝗶𝘁 𝗺𝗮𝗸𝗲 𝘀𝗲𝗻𝘀𝗲?

— Strategically: UniCredit would gain a much stronger foothold in Europe’s largest economy, complementing its Italian, Austrian, and CEE strength.

— Financially: There are potential cost synergies - branch consolidation, tech platform overlaps, funding advantages - but also major execution risks.

— Politically: Germany has historically resisted foreign takeovers of its banking champions. The key question now is whether Berlin ultimately blocks the deal - or accepts European banking consolidation.

𝗠𝘆 𝘁𝗮𝗸𝗲:

Europe often talks about the need for bigger, stronger banks. On paper, UniCredit–Commerzbank would fit that logic: UniCredit brings profitability and efficiency, Commerzbank offers strategic positioning in Germany. But in practice, politics make a full takeover complex - even if UniCredit’s latest move shows it is willing to keep increasing pressure.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) Mastercard buys BVNK

Mastercard buys BVNK for $1.8 billion. This is the largest stablecoin-focused acquisition to date. Here’s why it’s important.

𝗪𝗵𝗮𝘁 𝗕𝗩𝗡𝗞 𝗱𝗼𝗲𝘀

• BVNK helps businesses send, receive, and hold stablecoins alongside traditional currencies through one platform.

• It provides the compliance, banking access, and real-time conversion needed to make stablecoin payments work like any other cross-border transaction.

• That’s what makes it interesting: BVNK has solved one of digital money’s hardest problems - linking blockchain assets with regulated payment networks.

• Its platform lets fintechs, merchants, and banks move liquidity instantly between stablecoins and fiat without managing crypto infrastructure or compliance.

𝗪𝗵𝘆 𝗠𝗮𝘀𝘁𝗲𝗿𝗰𝗮𝗿𝗱 𝘄𝗮𝗻𝘁𝘀 𝗕𝗩𝗡𝗞

• This is not a crypto play but it’s about Mastercard owning how stablecoins integrate into regulated payment flows, i.e. where digital assets meet banking infrastructure, compliance, and real life use cases.

• Stablecoins are already being used for faster cross-border settlement, internal treasury movements, and merchant payouts, especially where traditional rails are slow or costly. But today, they often sit outside the systems most businesses rely on.

• BVNK removes that friction by making stablecoins behave like any other payment method - with built-in compliance, banking access, and seamless conversion between fiat and digital currencies.

• The move allows Mastercard to extend its network beyond cards, so whether a transaction is funded by fiat, stablecoins, or a mix of both, it can still flow through Mastercard’s infrastructure.

• Instead of being bypassed by new rails, Mastercard can now sit in the flow handling conversion, routing, and settlement across currencies and networks.

𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• Stablecoins shift the profit pool. The value is moving into FX, conversion, and liquidity management – all areas around payments.

• Stablecoins introduce an alternative to correspondent banking. Expect this to gradually start showing up in pricing and deal structures.

• Focus is shifting to the decision making around the transaction, i.e. whether it goes through cards, A2A or stablecoins. This choice directly affects cost, speed, and economics.

• Mastercard’s move puts pressure on Visa by moving earlier into the layer where stablecoins are converted, routed, and settled. If Visa doesn’t match that, it risks relying on someone else’s infrastructure.

• For fintechs, this removes a big hurdle. Instead of building banking relationships in every country and separate crypto infrastructure, they can use platforms like this to move money globally from day one.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

4) State of Crypto 2026

The report is out: State of the Crypto Industry 2026. If you want to understand where the industry is heading, here are my main takeaways.

𝟭. Crypto platforms are being evaluated on their ability to run compliant, reliable systems at scale, where onboarding, monitoring, and transaction controls are expected to work together consistently under regulatory pressure - and not as separate functions.

𝟮. Identity verification is no longer a one-off check but a continuous capability embedded across the full user lifecycle, influencing risk decisions, conversion, and regulatory outcomes.

𝟯. Performance improvements are coming from how verification flows are structured, with platforms reducing unnecessary steps, limiting retries, and routing users based on risk.

𝟰. Fraud activity is becoming more organized and persistent, with automated attacks, synthetic identities, and coordinated behavior targeting weaknesses in verification and monitoring, requiring systems that can detect patterns and isolated events.

𝟱. AI is being applied to connect identity data, user behavior, and transaction activity into a single decision process, allowing platforms to identify risks in real time and adjust controls dynamically.

𝟲. Most platforms are combining internal decision-making with external verification capabilities, keeping control over risk logic and user experience while relying on specialized providers for document verification, biometrics, and screening.

𝟳. Regulatory frameworks such as MiCA and the Travel Rule are directly affecting how products are built and how transactions are processed, with many firms still facing challenges in implementation, data sharing, and cross-border requirements.

𝟴. Transaction value is increasingly driven by businesses rather than individuals, reflecting the growing use of crypto for treasury management, cross-border settlement, and day-to-day movement of funds.

𝟵. Stablecoins are increasingly used as a transaction medium where stability and efficiency matter, shifting activity toward payment and settlement use cases rather than trading.

𝟭𝟬. Platforms are redesigning onboarding so that the level of verification matches the level of risk, reducing unnecessary steps for low-risk users while applying deeper checks only where needed, as both over-checking and under-checking create direct commercial and regulatory costs.

𝗢𝗻𝗲 𝘁𝗵𝗶𝗻𝗴 𝗶𝘀 𝗰𝗹𝗲𝗮𝗿:

As crypto matures, the focus is moving away from simply enabling access to assets toward managing how transactions are executed in practice, including who is allowed to participate, how those decisions are made, and how risk and compliance are applied throughout the lifecycle

Opinions: Panagiotis Kriaris, Source: Sumsub

Link to the report: https://sumsub.com/crypto-industry-report-2026/

Subscribe now

5) ChatGPT Instant Checkout

A few months ago the idea of native checkout inside ChatGPT looked like the next big step in commerce. Not any more.

𝗜𝗻𝘀𝗶𝗱𝗲 𝗖𝗵𝗲𝗰𝗸𝗼𝘂𝘁 𝗶𝗻 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 was launched by OpenAI in September 2025, essentially allowing users to purchase products directly inside the chat interface.

• As product discovery increasingly moved to AI assistants, the next logical step was to allow the transaction to happen in the same interface.

• The idea was simple: if users already search, compare and decide inside ChatGPT, sending them to another website creates friction.

• Instant Checkout was therefore an attempt to turn ChatGPT from a discovery layer into a commerce interface.

• The goal for OpenAI was to capture more of the commerce journey by keeping users inside ChatGPT from product search all the way to payment.

𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱 𝗶𝗻 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗲:

• Adoption remained extremely limited, with only a small number of merchants integrating the native checkout capability.

• User behaviour showed that people were comfortable using ChatGPT to discover and compare products, but they were not completing purchases inside the chat interface.

Recent reports indicate that OpenAI scaled back its plans for direct checkout inside ChatGPT, moving away from the “buy inside the chat” model and redirecting purchases to merchants’ websites or apps.

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗲:

• Agentic commerce may emerge in stages rather than all at once. The first role AI assistants are clearly winning is discovery, comparison and decision support. The step where the agent actually executes the payment appears to be much harder to shift because it sits on top of trust, payment credentials, merchant integrations and fulfilment infrastructure.

• The interface alone is not enough to move the transaction layer. Even if an AI assistant becomes the place where purchase decisions are made, the payment still tends to happen where the merchant relationship, payment acceptance, logistics and customer support already exist.

• This could suggest that the early model for agentic commerce may look more like orchestration than direct execution. AI agents may guide the purchase journey and direct the transaction to existing merchant systems, rather than fully replacing them.

• For the ecosystem, this shifts the focus to integration. What will matter most is which platforms, merchants and payment providers make it easiest for AI agents to connect to their systems and complete transactions. Those that become the default connections for AI assistants could end up capturing a significant share of the value in this new model.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

Understanding payments is about understanding distribution - and how control over merchant access has shifted over time. Here is the evolution.

𝗕𝗮𝗻𝗸𝘀

Banks originally controlled payments because they controlled the merchant accounts and settlement access required to move money through card networks. Payment acceptance was essentially a banking product tied to acquiring relationships.

𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗼𝗿𝘀

As card volumes increased, banks outsourced the operational layer of authorization, clearing, and settlement to processors. Processors became the technology backbone of card payments while banks retained the licenses and merchant relationships.

𝗜𝗦𝗢𝘀

Processors and acquiring banks lacked scalable merchant acquisition. ISOs emerged as independent sales organizations specializing in merchant distribution, building sales networks to originate merchant accounts and bring volume to acquirers.

𝗣𝗦𝗣𝘀

The growth of e-commerce created a new bottleneck: integration. PSPs reduced the complexity of connecting to multiple acquirers, gateways, and payment methods by offering a single integration layer for online merchants.

𝗜𝗦𝗩𝘀

With the rise of SaaS, vertical software platforms became the operating systems of merchants’ core business activities (POS, reservations, ERP, booking, etc.). Because these platforms controlled the merchant workflow, they became the most efficient channel for distributing payment acceptance.

𝗣𝗮𝘆𝗙𝗮𝗰𝘀

Once software platforms controlled merchant distribution, traditional acquiring onboarding became inefficient. PayFac models allowed platforms to aggregate merchants under a master merchant account, simplifying onboarding and enabling participation in payment economics.

The most important shift in payments today is happening at the software layer.

For the first time, merchant distribution is controlled by software platforms.

At this layer, payments are delivered through two models: ISVs and PayFacs. Both sit at the platform that owns the merchant relationship - but take different approaches.

𝗜𝗦𝗩 𝗺𝗼𝗱𝗲𝗹

The platform integrates payments into its software, but merchants retain their own merchant accounts with a PSP or acquirer. The platform distributes payments but remains outside the money flow.

𝗣𝗮𝘆𝗙𝗮𝗰 𝗺𝗼𝗱𝗲𝗹

The platform aggregates merchants under its own master merchant account and onboards them as sub-merchants. The platform becomes part of the payment flow and captures part of the economics.

The distinction between the two models is increasingly blurred:

• Many platforms start as ISVs, distributing payments through PSP partnerships.

• As transaction volume grows, they shift toward PayFac models to control onboarding and economics.

• At that point, the platform effectively becomes a payment provider.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) Agentic Commerce Market Map

Want to understand agentic commerce? This is a breakdown of the emerging stack and who does what.

𝟭. 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹𝘀

Models such as OpenAI, Anthropic, Meta, xAI provide the reasoning layer that allows agents to interpret instructions, plan actions and make decisions.

Without this layer, there are no autonomous agents.

𝟮. 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲

Providers such as AWS, Google Cloud, Cloudflare, Akash supply the compute and networking needed to run models and agents continuously.

This is the infrastructure layer of the agent economy.

𝟯. 𝗔𝗴𝗲𝗻𝘁 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀

Frameworks like MCP and A2A allow developers to build agents that can call APIs, access services and coordinate tasks.

This layer enables models to operate as agents.

𝟰. 𝗔𝗴𝗲𝗻𝘁 𝗻𝗲𝘁𝘄𝗼𝗿𝗸𝘀

Protocols such as Virtuals Protocol, Bittensor or Heurist allow agents to collaborate and coordinate with other agents rather than operating individually.

These networks provide shared environments where agents can exchange tasks and services.

𝟱. 𝗗𝗶𝘀𝗰𝗼𝘃𝗲𝗿𝘆

Before an agent can act, it must discover available services, APIs or resources.

Tools like x402scan and Unicity Labs allow agents to discover APIs, services or payment endpoints across the ecosystem.

𝟲. 𝗜𝗱𝗲𝗻𝘁𝗶𝘁𝘆 & 𝘁𝗿𝘂𝘀𝘁

Agents must prove who they are and whether they can be trusted.

Protocols such as ERC-8004, Cred Protocol, AgentProof provide identity and and verifiable credentials so agents can transact securely.

𝟳. 𝗙𝗮𝗰𝗶𝗹𝗶𝘁𝗮𝘁𝗼𝗿𝘀

Platforms like Stripe, Coinbase, Openx402, thirdweb connect agents to services, payments and workflows.

They act as the execution layer that lets agents actually do things.

𝟴. 𝗪𝗮𝗹𝗹𝗲𝘁𝘀 & 𝗮𝗰𝗰𝗼𝘂𝗻𝘁 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲

Solutions such as Privy, MetaMask, Fireblocks, Coinbase Wallet allow agents to hold assets, manage keys and sign transactions.

Technologies like ERC-4337 simplify account management so agents can transact programmatically.

𝟵. 𝗣𝗮𝘆𝗺𝗲𝗻𝘁 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲

Infrastructure such as x402, Stripe, Visa, Crossmint, Moonpay enables automated payments and settlement.

This is what allows agents to pay for services or receive payments automatically.

𝟭𝟬. 𝗕𝗹𝗼𝗰𝗸𝗰𝗵𝗮𝗶𝗻𝘀

Networks like Base, Solana, Polygon, Avalanche, Arbitrum provide the settlement and execution environment where transactions are recorded.

𝟭𝟭. 𝗦𝘁𝗮𝗯𝗹𝗲𝗰𝗼𝗶𝗻𝘀

Assets such as USDC and USDT provide programmable digital money that agents can move instantly across networks.

For many agent transactions, stablecoins act as the settlement asset.

𝟭𝟮. 𝗨𝘀𝗲𝗿 𝗶𝗻𝘁𝗲𝗿𝗳𝗮𝗰𝗲𝘀

Interfaces such as ChatGPT, Claude or Gemini are becoming the entry point where humans interact with agents and delegate tasks.

These interfaces increasingly act as the control layer for agent activity.

Opinions: Panagiotis Kriaris, Graphic source: Artemis Analytics

Subscribe now

3) The Invisible Bank

Banking’s next transformation is not about channels, apps or interfaces. The bank of the future may be the one we never see. The invisible bank is coming.

The shift has been unfolding via 3 distinct phases:

𝗣𝗵𝗮𝘀𝗲 𝟭: Branches – Banking was tied to location. You went to a branch. Products were standardized. The bank was a place you visited.

𝗣𝗵𝗮𝘀𝗲 𝟮: Mobile – Banking moved to the phone. Apps replaced branches, but customers still had to log in, search, and interact with the bank directly.

𝗣𝗵𝗮𝘀𝗲 𝟯: Invisible – Now banking is moving into the background. Services are triggered by context, decisions are increasingly supported by AI, and instead of customers going to their bank, the bank goes to them.

5 main transformation blocks are driving the shift:

𝟭. Interactions shifting from apps and interfaces to natural language conversations.

𝟮. Personalisation moving from a premium service to a mass capability.

𝟯. Work increasingly organised around hybrid human–AI teams.

𝟰. Decision making shifting from static, rule-based processes to intelligent systems.

𝟱. Banking systems being rebuilt as AI-native architectures.

In the future, there will be two types of institutions: those built around AI - and those merely using it.

The difference will determine who manages the transition and who struggles.

At the center of this shift lies 𝗮 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Will AI sit on top of the bank as another layer?

OR

Will the bank itself be rebuilt around intelligence?

If the latter is true, then the real question is practical: 𝗵𝗼𝘄 𝗱𝗼 𝗯𝗮𝗻𝗸𝘀 𝗰𝗮𝗻 𝗺𝗮𝗸𝗲 𝘁𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻?

Huawei has one of the clearest transformation blueprints in the industry. They call it Intelligent Finance Value Implementer and is built around 3 pillars:

𝟭. 𝗦𝗰𝗲𝗻𝗮𝗿𝗶𝗼𝘀:

AI starts with where the bank needs results and delivers value across 3 areas:

• Customer experience: interactions handled through conversational interfaces

• Risk & control: credit assessments, fraud monitoring, compliance checks via automated analysis

• Internal efficiency: operational workflows supported by AI assistants

𝟮. 𝗦𝘆𝘀𝘁𝗲𝗺 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴:

Many AI projects fail because they remain isolated pilots that never scale.

System engineering means building AI properly into the bank’s systems from the beginning - connecting data, security, governance, and infrastructure.

𝟯. 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲:

AI changes the technical demands on banking systems.

Banks need architectures combining general computing, AI computing, storage, and networking into a unified stack supporting both traditional workloads and AI-driven operations.

As AI becomes embedded across processes, systems must process data instantly, handle heavy model workloads, and remain continuously available.

Opinions: Panagiotis Kriaris, Graphic source: Huawei

Subscribe now

4) Top GenAI Platforms

Where is generative AI being used today - and what does that tell us about where the market is heading?

a16z has just published a list of the top 100 GenAI consumer apps - this is my summary of the ranking and what’s behind.

𝗛𝗼𝘄 𝘁𝗼 𝗿𝗲𝗮𝗱 𝘁𝗵𝗲 𝗿𝗮𝗻𝗸𝗶𝗻𝗴𝘀:

• ChatGPT leads in both web traffic and mobile users, reinforcing its position as the default entry point for consumer AI.

• The web ranking is concentrated around AI assistants and AI search tools - ChatGPT, Gemini, Perplexity, Claude and Grok - competing to become the main AI user interface.

• Mobile usage is driven primarily by creative and editing applications such as CapCut, Remini, PicsArt and Canva, where AI is used directly inside photo, video and design workflows.

• Large software platforms that have integrated AI features rank high, showing how AI adoption often happens inside tools people already use.

• Multiple Chinese apps rank among the top mobile GenAI products (Doubao, Meitu, Baidu AI Search, QQ Browser), reflecting the scale of China’s domestic AI market.

• Platforms used by developers to build AI applications attract significant traffic.

𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗰 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• A small number of AI assistants are becoming the main entry point to AI. ChatGPT, Gemini, Claude and Grok are competing to define the primary interface for accessing models, tools and services.

• Distribution is becoming the decisive advantage. Platforms with large installed user bases (Canva, Notion, CapCut, Grammarly) can scale AI usage rapidly because AI is embedded directly inside workflows people already use.

• AI usage is fragmenting by task rather than converging around a single dominant product. Search, coding, design, voice generation and content creation are already emerging as distinct product categories.

• Consumer adoption is driven by content creation. Photo, video and design tools dominate mobile AI usage, highlighting visual creation as one of the first mass use cases of generative AI.

• China has developed a large domestic AI app ecosystem, with platforms such as Doubao, Meitu and Baidu AI Search reaching major user scale.

• Developer platforms are attracting substantial traffic because every new AI product starts there. Tools such as Hugging Face, Cursor and Google AI Studio are becoming central infrastructure for building AI applications.

• The AI market is evolving as a layered ecosystem. Interfaces, AI embedded in software platforms, specialised applications and development tools each capture different parts of the value chain.

𝗪𝗵𝗮𝘁’𝘀 𝗻𝗲𝘅𝘁:

The next phase of competition will be about distribution. The companies that control the interface, the workflow or the developer tools will shape how AI reaches users.

Opinions: my own, Graphic and data source: a16z speedrun

Subscribe now

Revolut has applied for a US bank charter. But what does it take to win in one of the toughest markets for challengers?

𝗪𝗵𝘆 𝘁𝗵𝗲𝘆 𝗰𝗵𝗼𝘀𝗲 𝗮 𝗨𝗦 𝗯𝗮𝗻𝗸 𝗰𝗵𝗮𝗿𝘁𝗲𝗿:

• Ability to operate across all 50 states under one regulatory framework.

• Full ownership of the customer experience, enabling faster product development and direct service delivery.

• Direct access to payment rails such as Fedwire and ACH, improving speed, reliability, and cost efficiency.

• Ability to offer loans and credit cards directly, capturing net interest margin (NIM) and core banking revenues.

• Ability to offer insured deposits to US customers.

For years, Revolut explored acquiring a US bank as a faster entry strategy. Here is 𝘄𝗵𝘆 𝘁𝗵𝗲𝘆 𝗳𝗶𝗻𝗮𝗹𝗹𝘆 𝗱𝗲𝗰𝗶𝗱𝗲𝗱 𝗮𝗴𝗮𝗶𝗻𝘀𝘁 𝗶𝘁:

• Allows the bank to be designed from scratch - technology stack, operating model, risk framework.

• Regulators would still need to approve the transformation of the bank, so buying one does not avoid regulatory scrutiny.

• Avoids the complex integration that often follows bank acquisitions across systems, culture, and compliance.

• Starts with a clean balance sheet and structure instead of restructuring an existing bank.

𝗧𝗵𝗲 𝗨𝗦 𝗺𝗮𝗿𝗸𝗲𝘁 𝗹𝗮𝗻𝗱𝘀𝗰𝗮𝗽𝗲:

• Thousands of banks exist due to historical state-based charters, yet a few large institutions dominate national market share.

• Most fintechs are not banks. They operate as front-end apps while regulated partner banks hold deposits, issue cards, and provide deposit insurance.

• Obtaining a banking license is slow, costly, and complex, which is why only a handful of digital banks operate with full charters.

• Payments infrastructure is not unified. Transfers and payments run across multiple rails (cards, ACH, real-time networks), which increases complexity for new entrants.

• Higher interchange fees make card spending a core revenue source.

• The market is credit-driven. Competing at scale requires lending capabilities, which brings regulatory and risk management challenges.

𝗪𝗵𝗮𝘁 𝗶𝘁 𝘁𝗮𝗸𝗲𝘀 𝘁𝗼 𝘄𝗶𝗻:

• Accounts alone don’t scale in the US. Real economics sit in credit, so Revolut will need strong lending and balance-sheet capabilities.

• Distribution matters as much as product. Growth will require partnerships, ecosystems, and embedded channels rather than relying on the app alone.

• Customer acquisition is expensive. Aggressive bonuses, rewards, and marketing make growth costly, so retention and revenue per customer are critical.

• Focus on segments where incumbents are weaker, such as international users and globally mobile customers where Revolut already has an advantage.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What is the one factor that will determine whether Revolut succeeds in the US market?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) API Marketplaces in Banking

It is one of the biggest misconceptions in the transition from app-driven to intelligent banking: APIs are no longer a focus…

In reality, they matter more than ever.

Banks are in the middle of a shift. They must move from a product-and-channel business model to a capability-and-platform one.

𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗺𝗼𝗱𝗲𝗹:

Value is created through owned products and channels, with APIs primarily used to connect front ends to internal systems.

𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝘁 𝗺𝗼𝗱𝗲𝗹:

Value is created through programmable capabilities embedded across external journeys and AI-driven workflows, with APIs acting as the commercial interface of the bank.

𝗧𝗵𝗲 𝗔𝗣𝗜 𝗠𝗮𝗿𝗸𝗲𝘁𝗽𝗹𝗮𝗰𝗲:

If APIs become the commercial interface of the bank, they need a commercial operating layer.

How do they do that?

Via an API marketplace.

This is essentially a digital storefront where banks can publish their APIs so partners and developers can easily find, test, and use them.

What does it do?

• Brings all APIs into one place

• Makes APIs easy to discover through a searchable catalogue

• Shows what each API does, how to use it, and what it costs

• Lets partners test APIs in sandbox environments

• Manages who gets access, under which plan, and with what limits

• Enables pricing, packaging, and monetisation of APIs

• Supports billing and payment collection

Building an API Marketplace is not just about publishing APIs.

It requires discovery, access control, sandboxing, developer experience, billing, and monetisation working together as one product.

𝗧𝗵𝗶𝘀 𝗶𝘀 𝘄𝗵𝘆 𝗺𝗼𝘀𝘁 𝗯𝗮𝗻𝗸𝘀 𝗰𝗮𝗻𝗻𝗼𝘁 𝗱𝗼 𝗶𝘁 𝗮𝗹𝗼𝗻𝗲.

Trying to build it internally is slow, complex, and expensive - and often results in a basic developer portal, not a true marketplace.

Expert providers like Digitalapi.ai can deliver a ready-made, white-label, proven marketplace layer that plugs into existing architecture and enables banks to launch and scale API monetisation quickly.

𝗪𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝗺𝗲𝗮𝗻 𝗶𝗻 𝗽𝗿𝗮𝗰𝘁𝗶𝘀𝗲?

• APIs are presented in one place with clear descriptions, documentation, and examples, so external teams can easily understand what exists and how to use it.

• Partners can request access, test APIs, and move to production through a structured self-service flow, removing long email chains and manual onboarding steps.

• APIs can be grouped into products, assigned pricing and usage plans, and billed automatically, enabling direct monetisation without custom commercial or technical builds.

Banks need to stop treating APIs as technical plumbing.

They need to recognise them as a core business enabler.

In the era of intelligent banking, the winners will not be the banks with the most APIs, but the ones that know how to package, distribute, and monetise their capabilities through them.

Opinions: Panagiotis Kriaris, Graphic sources: DigitalAPI, Panagiotis Kriaris

Subscribe now

3) Stablecoin Issuance Models

Everyone is talking about stablecoins. But they’re not all built the same - issuers, business models, and strategies can differ widely. Here’s an overview.

𝗧𝗵𝗲 𝗶𝘀𝘀𝘂𝗲𝗿𝘀

- Specialized fintechs / crypto-native firms → e.g. Tether, Circle, Paxos. These dominate today, building their business model around issuance and reserve management.

- Payment companies & platforms → e.g. PayPal (PYUSD). They issue stablecoins as an extension of their ecosystem to drive payments, retention, and new financial services.

- Banks → from global banks like JPMorgan (JPM Coin) to regional or central banks exploring tokenized deposits or settlement coins. Banks may issue directly for wholesale settlement or customer-facing stablecoins.

- Big Tech / platforms → e.g. Meta’s abandoned Diem project. Tech platforms see stablecoins as a way to lock users into closed ecosystems.

- Central Banks (CBDCs) → officially distinct from stablecoins, but they share the core principle of fiat-backed digital tokens issued and redeemed 1:1.

𝗧𝗵𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗺𝗼𝗱𝗲𝗹𝘀

1. Direct issuance → The issuer mints (creates) and redeems (removes from circulation) coins itself, fully controlling reserves. Customers deposit fiat (e.g. USD), the issuer mints stablecoins, and the reserves are invested in highly liquid, low-risk assets (US Treasuries, repos, deposits at systemically important banks).

Example: Tether (USDT) and Circle (USDC).

Revenue model: Issuers earn interest on reserves. In 2024, Circle distributed ~60% of that income to partners like Coinbase, while Tether retains it all.

2. White-label issuance → The issuer mints on behalf of a partner, who owns the distribution relationship. The economics are often flipped: most reserve income goes to the partner, while the issuer charges a management fee.

Example: Paxos, issuing PayPal’s PYUSD and Global Dollar (USDG) for partners like Robinhood and Kraken.

Both models rely on the same reserve mechanics - but differ in who owns the customer relationship and who captures the economics.

𝗧𝗿𝗲𝗻𝗱𝘀

- Distribution is key. Issuance alone doesn’t drive adoption - that’s why Circle shares reserve income with Coinbase and Paxos passes most to PayPal.

- Tokenization is the growth driver. From powering crypto and xborder payments today, stablecoins can become the natural settlement layer for tokenized assets - equities, bonds, even mortgages.

- From reserves to transactions. Revenues could shift from interest on reserves to transaction fees, moving stablecoins closer to payments than balance-sheet plays.

- Banks + fintechs converge. As banks issue tokens, the line between stablecoins and tokenized deposits will blur — with regulation pushing them together.

- Diverging strategies. Tether keeps all income, Circle shares to grow, Paxos builds infrastructure via white-label.

Opinions: Panagiotis Kriaris, Graphic source: Goldman Sachs

Subscribe now

4) Agentic Commerce Use Cases

Everyone is talking about agentic commerce. But what will agents actually do?

So, I mapped out the use cases.

𝟭. 𝗕𝗶𝗹𝗹 & 𝗶𝗻𝘃𝗼𝗶𝗰𝗲 𝗽𝗮𝘆𝗺𝗲𝗻𝘁 agents: Automatically pay recurring bills, detect anomalies, and choose the optimal payment method without user involvement.

𝟮. 𝗣𝗿𝗶𝗰𝗲 𝗰𝗼𝗺𝗽𝗮𝗿𝗶𝘀𝗼𝗻 & 𝘀𝘄𝗶𝘁𝗰𝗵𝗶𝗻𝗴 bots: Continuously monitor prices across providers and automatically switch to better deals based on user-defined rules.

𝟯. 𝗖𝗼𝗻𝘁𝗿𝗮𝗰𝘁 & 𝘀𝘂𝗯𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻 optimisers: Track subscription usage and contract terms to cancel, downgrade, or switch providers to reduce waste and cost.

𝟰. 𝗣𝗿𝗼𝗰𝘂𝗿𝗲𝗺𝗲𝗻𝘁 agents: Autonomously source suppliers, compare quotes, negotiate terms, and place purchase orders within predefined constraints.

𝟱. 𝗜𝗻𝘃𝗲𝗻𝘁𝗼𝗿𝘆 & 𝗿𝗲𝗽𝗹𝗲𝗻𝗶𝘀𝗵𝗺𝗲𝗻𝘁 agents: Forecast stock levels, trigger restocking, and coordinate multi-supplier orders based on demand patterns.

𝟲. 𝗕𝘂𝗻𝗱𝗹𝗲 𝗰𝗼𝗺𝗽𝗮𝗿𝗶𝘀𝗼𝗻 agents: Assemble and compare multi-service bundles (e.g., telecom, insurance, SaaS) to find the most cost-effective package.

𝟳. 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗽𝗿𝗼𝗱𝘂𝗰𝘁 𝘀𝗲𝗹𝗲𝗰𝘁𝗶𝗼𝗻 agents: Translate a customer need (e.g. set up a home office, prepare for a trip) into a fully assembled, ready-to-buy product list or cart.

𝟴. 𝗟𝗼𝘆𝗮𝗹𝘁𝘆 & 𝗿𝗲𝘄𝗮𝗿𝗱𝘀 bots: Recommend where to buy or how to pay to maximise loyalty points, cashback, discounts, etc.

𝟵. 𝗠𝗶𝗰𝗿𝗼-𝘁𝗮𝘀𝗸 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗲 agents: Execute quick, low-value actions (e.g. mobile top-ups, parking payments, digital vouchers).

𝟭𝟬. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁𝘂𝗮𝗹 𝘀𝗵𝗼𝗽𝗽𝗶𝗻𝗴 copilots: Assist users in discovering the right products by filtering, comparing, and explaining options based on their preferences.

𝟭𝟭. 𝗖𝗵𝗲𝗰𝗸𝗼𝘂𝘁 & 𝗽𝗮𝘆𝗺𝗲𝗻𝘁 𝗿𝗼𝘂𝘁𝗶𝗻𝗴 agents: Optimise payment routing at checkout by applying offers, selecting the best instrument, or splitting payments.

𝟭𝟮. 𝗖𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 & 𝗳𝗼𝗿𝗺-𝗳𝗶𝗹𝗹𝗶𝗻𝗴 agents: Pre-fill complex forms (KYC, claims, applications) using existing data and guide the user through final validation steps.

As strange as it sounds, agents will add complexity before they reduce it:

• Agents introduce a new decision layer that evaluates options, applies constraints, and coordinates steps across independent systems.

• The result is a single, consolidated output at the experience level instead of a chain of fragmented, intermediate decisions.

Which use cases would you add?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

Why would Stripe buy a PayPal under pressure - and how would that change the industry?

𝗦𝘁𝗿𝗶𝗽𝗲 𝘃𝘀. 𝗣𝗮𝘆𝗣𝗮𝗹:

Stripe sits primarily on the merchant side of the stack, while PayPal at the consumer interaction layer.

• Stripe controls a large share of digital commerce infrastructure, powering APIs, orchestration, and embedded finance for millions of merchants - deeply embedded on the supply side but without a scaled consumer identity layer.

• Stripe growth comes from developer adoption and platform integrations.

• PayPal controls consumer authentication at checkout, with hundreds of millions of active accounts embedded across global e-commerce. It directly shapes payment method selection and affects merchant conversion rates at checkout.

• PayPal’s strength is distribution and consumer trust. Despite technical complexity, it maintains a persistent consumer relationship and strong brand recognition.

𝗪𝗵𝘆?

The transaction would reposition Stripe from infrastructure provider to platform player with influence across the full commerce flow.

• The acquisition would combine infrastructure with consumer identity, removing the current separation between merchant orchestration and wallet authentication. This gives Stripe greater control over pricing, routing, and payment method presentation at checkout.

• Stripe would move from a merchant-focused provider to a scaled two-sided platform with value coming from combining merchant integration with consumer distribution.

• The combined entity would gain end-to-end transaction data across both sides, strengthening fraud models, underwriting, and embedded financial services.

• It would reduce dependency on third-party wallets and limit the risk of disintermediation at checkout.

𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• The payments stack would become more vertically integrated, with one player controlling orchestration, wallet identity, and financial services. This would increase competitive pressure on standalone PSPs and niche providers.

• Merchant negotiating power would weaken as pricing and checkout positioning consolidate within a single platform. Large digital merchants would face fewer strategic alternatives.

• Wallet competition would intensify, pushing rivals toward stronger ecosystem lock-in strategies. The focus would shift from acceptance to platform control.

• Banks would lose further visibility at the point of transaction, remaining balance sheet providers unless they strengthen distribution and embedded strategies.

𝗕𝗼𝘁𝘁𝗼𝗺 𝗹𝗶𝗻𝗲:

Scale alone is not enough if you control only one side of the transaction - real leverage comes from controlling both infrastructure and customer identity.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) Revolut’s 2.0 Play

Revolut’s real success story isn’t the app, the products, or the UX. It’s how it became a bank without looking like one. Here’s how they did it.

𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗠𝗼𝗱𝗲𝗹:

• Revolut started as a single-product proposition focused on low-cost foreign exchange and card-based spending for internationally mobile customers driven by FX and premium subscriptions.

• The initial model was asset-light, cross-border, and growth-led: acquire users fast, expand geographically, and deepen engagement through UX and pricing transparency.

• It progressively expanded into adjacent financial services: savings, trading, crypto, insurance distribution, business accounts and payments. The app became a distribution platform.

• The revenue mix evolved from transaction-driven income to a diversified model including subscriptions, trading revenues, and increasingly, interest income from customer deposits.

• Today, the business resembles a digital financial group: multi-product, multi-entity, operating across jurisdictions with banking licences, holding material deposits and facing bank-level regulatory scrutiny.

𝗛𝗼𝘄 𝘁𝗵𝗲𝘆 𝗱𝗶𝗱 𝗶𝘁:

• Start with a high-frequency use case that drives repeat, embedded usage, then layer products into an engaged customer base.

• Single front-end, modular back-end: one interface distributing multiple products, enabling rapid rollout without fragmenting UX.

• Aggressive geographic expansion early on, creating scale before incumbents reacted.

• Strong data feedback loops: transaction data used to inform product design, pricing, risk, and targeted cross-sell.

• Tiered monetisation: free entry point to maximise reach, subscription tiers for ARPU, and event-driven monetisation (travel, trading, business).

• Relentless focus on product velocity: rapid launches, iteration, and testing new verticals ahead of traditional players.

• Capital strategy aligned with growth: significant equity funding to subsidise expansion until profitability became sustainable.

𝗧𝗵𝗲 𝟮.𝟬 𝗣𝗹𝗮𝘆:

• Shift to becoming the customer’s primary bank: salary account, savings, and credit relationship.

• Focus on deposit consolidation and balance-sheet depth.

• More bank-like revenue mix driven by net interest margin and lending alongside fees and subscriptions.

• Institutional-grade governance, controls, risk management, and regulatory focus.

• Greater exposure to rate cycles, asset-liability management, and credit risk.

• Operational resilience, AML, fraud, and compliance.

• Geographic complexity: different licences, regulators, capital requirements, and local dynamics.

• Main challenge: preserving product speed and innovation culture within a heavily regulated banking framework.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What do you think is the single most important lesson from this journey?

Opinions: Panagiotis Kriaris, Graphic source: C-Innovation, Revolut’s Expansion Strategy: From Neobank to Institutional Powerhouse. https://www.c-innovation.eu/

Subscribe now

3) Banks’ Ecosystem Play

It is one of the most widespread myths in banking: banks cannot play the ecosystem game.

I am not sure when I heard this for the first time, but it’s becoming increasingly hard to defend.

Not because banks have suddenly become good at building ecosystems (most haven’t), but because the reality on the ground - especially when you look at the breadth of tools and capabilities now readily available - tells a very different story.

𝗘𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺𝘀 𝗮𝗿𝗲 𝗻𝗼 𝗹𝗼𝗻𝗴𝗲𝗿 𝗮 𝗰𝗵𝗼𝗶𝗰𝗲 𝘁𝗼 𝗺𝗮𝗸𝗲 𝗼𝗿 𝗻𝗼𝘁 – they have become the operating environment of the modern economy:

• Customers already live inside ecosystems (BigTech, platforms, Superapps). Banks are often just one service inside someone else’s flow.

• Value is created across journeys, not products - payments, lending, identity, data, all blend into a single experience.

• Distribution power has shifted. Whoever owns the interface and the data sets the rules - everyone else competes on margins.

• Standalone propositions are being eroded. Even strong offerings lose relevance if they are not embedded in a broader context.

This is exactly why 𝗼𝗽𝘁𝗶𝗻𝗴 𝗼𝘂𝘁 𝗶𝘀 𝗻𝗼𝘁 𝗮 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗳𝗼𝗿 𝗯𝗮𝗻𝗸𝘀.

• If you don’t own distribution, you become invisible infrastructure in someone else’s experience.

• Product excellence is no longer a guarantee for value - control of the customer journey has become more important.

• Even regulation is no longer the moat it used to be. In some cases, it accelerates disintermediation (think open banking access)

• Disintermediation doesn’t necessarily require disruption. You don’t need to be replaced to be sidelined - you just need to lose the primary relationship.

𝗔𝗹𝗹 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗹𝗲𝗮𝗱𝘀 𝗺𝗲 𝘁𝗼 𝘁𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝘀𝘁 𝗺𝗶𝘀𝗰𝗼𝗻𝗰𝗲𝗽𝘁𝗶𝗼𝗻 𝗼𝗳 𝗮𝗹𝗹:

playing the ecosystem game does not mean banks must become BigTech.

𝗕𝘂𝘁 𝗶𝘁 𝗱𝗼𝗲𝘀 𝗿𝗲𝗾𝘂𝗶𝗿𝗲 𝗮 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗺𝗶𝗻𝗱𝘀𝗲𝘁:

• Stop trying to own everything. Orchestrating third-party products can work extremely well.

• Banks don’t need to own the whole stack to win. They need to own a critical role in the value chain and be indispensable to the end outcome.

• Composable architecture is no longer nice to have - APIs, event-driven systems, and cores that plug directly into external journeys are a prerequisite.

• Choose your partners carefully. Most ecosystem value is captured (or lost) at the integration layer.

• Redefine success: from product P&Ls to distribution reach, embedded transaction volume, and frequency of use across external journeys.

So, 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗯𝗮𝗻𝗸𝘀 is not whether ecosystems make sense.

𝗕𝘂𝘁 𝗿𝗮𝘁𝗵𝗲𝗿:

Are banks willing to put in the effort to become part of the game – and do they really understand the cost of staying out?

Opinions: Panagiotis Kriaris, Graphic sources: Celent, Mambu

Subscribe now

4) Stablecoins Report

The report is out. A practical deep dive into how stablecoins are becoming embedded in global financial infrastructure.

📘 Stablecoins: The Operating Layer for Global B2B Payments 2026 Report

I’m pleased to co-publish this together with Venturebloxx and Finance Loop – Alliance for On-Chain Finance, alongside Fintech Wrap Up, London Blockchain Conference, Januar, Utila, Range, and Depa.

Global payments today are no longer about choosing a single rail, but about intelligently orchestrating multiple rails based on:

• Corridor economics
• Liquidity timing and cut-offs
• Compliance and regulation
• Reconciliation and audit requirements

Stablecoins are not replacing traditional payment rails; they are integrating into them and changing the orchestration logic of global value transfer.

As adoption matures, they will increasingly function as invisible infrastructure - embedded within banking, PSP, and enterprise payment offerings.

The numbers:

• ~$27T+ annualised stablecoin settlement volume

• ~$300–310B global fiat-backed stablecoin market cap

• ~$16.5T–$23.7T stablecoin-addressable B2B payments TAM

• Stablecoins ≈30% of all on-chain transaction activity

• Euro stablecoins exceeded €500M market cap

• LatAm payment corridors saw ~9× growth in stablecoin flows.

👉 Download the report here: https://venturebloxx.com/stablecoin-b2b-payments

Subscribe now

5) Digital Finance Trends

Stripe’s annual newsletter is out. The most interesting part isn’t Stripe itself, but the structural shifts it highlights. Here’s my summary.

𝟭. 𝗠𝗮𝗿𝗸𝗲𝘁𝘀 structurally concentrating
Profits, capital, and talent are concentrating faster than ever before, with top firms capturing a disproportionate share of profits and market value - widening the gap between leaders and the rest.

𝟮. 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲, 𝗰𝗼𝗺𝗽𝘂𝘁𝗲, 𝗮𝗻𝗱 𝗱𝗮𝘁𝗮 infrastructure as the primary engines of economic growth
Nearly half of US GDP growth is driven by software, computers, and data center investment signaling a long-term shift of economies toward digital and AI-native production.

𝟯. 𝗘𝗻𝘁𝗿𝗲𝗽𝗿𝗲𝗻𝗲𝘂𝗿𝘀𝗵𝗶𝗽 entering a faster pace
Company formation, product launches, and code production are all accelerating sharply, with startups reaching meaningful revenue milestones much faster than in the past - suggesting it is becoming much easier to build and scale companies.

𝟰. 𝗚𝗹𝗼𝗯𝗮𝗹 𝗯𝘆 𝗱𝗲𝗳𝗮𝘂𝗹𝘁 as the standard operating model
New digital businesses increasingly treat the entire internet as their domestic market, with meaningful revenue coming from countries far outside traditional Tier-1 economies which means global payments, tax, and compliance have to be in place from day one.

𝟱. 𝗦𝘁𝗮𝗯𝗹𝗲𝗰𝗼𝗶𝗻𝘀 becoming payment rails
Stablecoin payment volume is decoupling from crypto speculation and growing rapidly, with most volume representing B2B activity - pointing to stablecoins as a practical cross-border settlement layer rather than a niche crypto asset.

𝟲. 𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 converging with programmable software infrastructure
Wallets, issuing, cards, billing, and treasury are increasingly API-native, composable, and embedded - blurring the line between fintech and general software platforms.

𝟳. 𝗚𝗿𝗼𝘄𝘁𝗵 increasingly depends on capital access
Traditional SME lending is down. Companies with access to data-driven, revenue-linked financing are growing faster making capital access increasingly decisive for growth.

𝟴. 𝗥𝗲𝘃𝗲𝗻𝘂𝗲 is increasingly infrastructure-driven
Capabilities such as optimization, localized checkout, alternative payment methods, and AI-driven authorization and fraud controls are directly increasing revenue -turning payments infrastructure into a primary growth driver rather than a cost center.

𝟵. 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗲 as a new channel
As purchasing shifts from human-driven browsing to software-driven execution, the competitive advantage will shift from UX and interface design to machine-readable identity, authentication, and payment infrastructure.

𝟭𝟬. 𝗥𝗲𝗴𝘂𝗹𝗮𝘁𝗶𝗼𝗻 as the biggest dependency on productivity gains
As technological progress is accelerating, productivity gains will depend on whether regulation can support faster adoption of new technologies.

Subscribe now

2025 full-year results. How do Visa and Mastercard compare and what’s behind the numbers?

𝗩𝗶𝘀𝗮 𝘃𝘀. 𝗠𝗮𝘀𝘁𝗲𝗿𝗰𝗮𝗿𝗱 𝗙𝗬 𝟮𝟬𝟮𝟱 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• Growth: Mastercard grew faster (+16% vs +11%) thanks to stronger gains in services; Visa’s growth came mainly from transaction volume.

• Revenue mix: Visa stays scale-driven; Mastercard’s mix is shifting toward data and security, helping earnings when spending slows.

• X-border: Both strong, but Mastercard (+15%) captured more travel recovery than Visa (+13%).

• Profitability: Visa held margins via cost control; Mastercard lifted them through higher-margin services.

• Geography: Visa’s strength is U.S. spending; Mastercard benefits more from international flows.

• Strategy: Visa leads on scale and efficiency; Mastercard’s growth reflects faster expansion beyond core payments.

𝗜𝗻 𝘀𝘂𝗺𝗺𝗮𝗿𝘆: Visa concentrates on scaling transaction volume and efficiency, while Mastercard is building more revenue streams around identity, security, and data services on top of the card network.

𝗖𝗼𝗻𝗰𝗹𝘂𝘀𝗶𝗼𝗻𝘀 𝗮𝗻𝗱 𝘁𝗿𝗲𝗻𝗱𝘀:

𝟭. Consumer spending remains steady despite higher rates, showing payments are less exposed to short-term swings.

𝟮. Both schemes are increasingly balancing transaction volumes with service-led growth, with growth now driven by technology, data, and services.

𝟯. Cross-border growth shows travel and global commerce remain the main sources of high-margin revenue.

𝟰. Cash keeps declining.

𝟱. Service-led growth is becoming central across the industry as transaction margins compress.

𝟲. Scale is still the schemes’ biggest strength — more volume doesn’t mean higher costs, which protects profitability.

𝟳. Diversification beyond transaction revenue is emerging as the new playbook for sustainable growth.

𝟴. Both are positioning for the next phase — AI for intelligence, stablecoins for speed, and agentic commerce for autonomous finance.

𝟵. Payments are entering a mature phase — growth is steady, adoption is fully digital, and new value is coming from data-driven services.

𝟭𝟬. Revenue is compounding faster than underlying volumes, signalling structurally improving monetisation per transaction.

𝟭𝟭. Tokenisation is becoming a material operating layer, strengthening security, acceptance, and network stickiness.

𝟭𝟮. Competition is shifting beyond pricing and toward distribution economics (incentives, issuer partnerships, fintech access).

𝗞𝗲𝘆 𝘁𝗮𝗸𝗲𝗮𝘄𝗮𝘆:

Competition is moving from pure transaction scale to expanding revenue per transaction through services, data, security, and embedded capabilities.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Which model has the better chance of winning long-term: Visa’s rails optimization or Mastercard’s rails monetization?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) Banks’ Visibility Gap

Payments have a visibility problem.

Millions of events, logs, transactions, alerts, and customer interactions scattered across dozens of systems that were never designed to provide a single, end-to-end view.

They can’t be searched.

They can’t be connected.

And they can’t be used in real time.

For banks, this means they only ever see the customer in hindsight - not at the moment when it matters.

And now with real-time payments (already) being the norm, that visibility gap becomes critical.

Banks now need to handle in seconds incidents that used to be managed in hours.

This is a fundamental operational shift.

And the biggest gap isn’t data.

𝗜𝘁’𝘀 𝘁𝗶𝗺𝗶𝗻𝗴.

To stay competitive in payments, banks don’t need more data lakes.

They need real-time visibility across the data they already have.

𝗦𝗼 𝗵𝗼𝘄 𝗱𝗼 𝘆𝗼𝘂 𝗳𝗶𝘅 𝘁𝗵𝗮𝘁?

Bank payment architectures were never built for real-time decisions. Most run on technology choices made 10, 20, even 30 years ago.

On the one hand, banks can’t replace their payment engines, fraud tools or reconciliation layers overnight.

On the other hand, delays that were once tolerable are no longer operationally viable.

What banks actually need is a real-time operational layer that sits on top of existing payment systems - making payment events searchable, connected, and observable end-to-end, without touching the underlying rails.

A new breed of technology providers are already addressing this market gap.

Elastic is a great example:

• It provides real-time visibility across payment flows by collecting transactions, events, logs, traces, and alerts from payment engines, fraud systems, APIs, and customer channels – and correlating them as they happen.

• It doesn’t replace cores, payment hubs, or scheme connectivity.

• It connects them – allowing banks to detect failed payments, fraud patterns, latency issues, and customer impact in real time.

𝗧𝗵𝗶𝘀 𝗶𝘀 𝗮 𝗽𝗼𝘄𝗲𝗿𝗳𝘂𝗹 𝗽𝗿𝗼𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻: not changing the systems banks rely on, but making them actionable at the moment decisions are made.

Real-time visibility isn’t just an efficiency upgrade.

It is the foundation that every real-time, AI-driven workflow, decision, and customer interaction will depend on.

𝗧𝗵𝗲 𝗿𝗲𝗮𝗹 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

How fast can banks move to a proactive approach if their visibility is still built for a reactive world?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) Stripe vs. Adyen

It’s one of the most popular comparisons in payments. But it’s also one of the most oversimplified. Stripe and Adyen are solving different problems.

𝟭) 𝗖𝗼𝗿𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀:

• Stripe builds programmable financial infrastructure: payments as a foundation, extended by billing, issuing, treasury, identity, and embedded finance capabilities.

• Adyen builds a unified payments and acquiring platform: one system that handles gateway, risk, processing, and local acquiring globally.

𝟮) 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗺𝗼𝗱𝗲𝗹:

• Stripe monetizes by expanding the number of financial services a customer uses.

• Adyen monetizes by processing very large volumes efficiently on a single stack.

𝟯) 𝗡𝘂𝗺𝗯𝗲𝗿𝘀:

• Both process roughly $1.4T in annual payment volume.

• Stripe generates significantly higher net revenue than Adyen, implying higher effective take rates and broader monetization beyond pure acquiring.

• Adyen generates much higher disclosed margins, reflecting operational efficiency and tight cost control.

• Stripe’s higher valuation comes from generating more revenue per dollar of volume via multiple products on top of payments. Adyen’s lower valuation reflects a model focused on margin and efficiency rather than revenue expansion.

𝟰) 𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗯𝗮𝘀𝗲:

• Stripe’s focus is SaaS companies, platforms, marketplaces, and digital-first businesses.

• Adyen’s focus is with large multinational merchants and global retailers.

𝟱) 𝗖𝗵𝗮𝗻𝗻𝗲𝗹 𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻𝗶𝗻𝗴:

• Stripe is dominant in online and embedded payment flows, with in-person as an extension.

• Adyen positions online and in-store as one unified environment.

𝟲) 𝗞𝗲𝘆 𝘀𝘁𝗿𝗲𝗻𝗴𝘁𝗵𝘀:

• Stripe: developer experience, speed of integration, breadth of APIs, product speed.

• Adyen: reliability, performance at scale, global acquiring coverage, unified data model.

𝟳) 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗰 𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻𝗶𝗻𝗴:

• Stripe has built a broad financial services platform on top of payments.

• Adyen is focused on building a tightly integrated global acquiring engine with strong operational control.

𝗦𝗼, 𝘄𝗵𝗶𝗰𝗵 𝗼𝗻𝗲 𝗶𝘀 𝗯𝗲𝘁𝘁𝗲𝗿?

These are two different models solving different problems.

Stripe’s bet is that value in payments shifts upward - into software layers, embedded finance, and monetizing financial services beyond pure transaction processing.

Adyen’s bet is that value stays concentrated in scale, reliability, and operating discipline.

Which one is better depends on which bet pays off more in the long term.

𝗪𝗵𝗶𝗰𝗵 𝗺𝗼𝗱𝗲𝗹 𝗶𝘀 𝘄𝗶𝗻𝗻𝗶𝗻𝗴?

Strangely enough both - but in different parts of the market.

Platforms and software companies are embedding payments to monetise financial flows (Stripe), while large global merchants are consolidating providers to reduce complexity and improve global optimisation (Adyen).

Opinions and graphics: Panagiotis Kriaris

Subscribe now

4) How to optimize content

If you write content, then you have to pay attention. LinkedIn has just published the do’s and don’ts for staying visible.

In the future, content that does not show up inside AI-generated answers will effectively not exist.

Here is my summary of the Linkedin guide:

𝟭. 𝗖𝗼𝗻𝘁𝗲𝗻𝘁 𝗺𝘂𝘀𝘁 𝘀𝗲𝗿𝘃𝗲 𝘁𝘄𝗼 𝗮𝘂𝗱𝗶𝗲𝗻𝗰𝗲𝘀 𝗮𝘁 𝗼𝗻𝗰𝗲: human readers and AI models. Plain, accessible language, a neutral authoritative tone, and sections that make sense on their own increase the chance content is trusted, quotable, and surfaced in AI-generated answers.

𝟮. 𝗖𝗹𝗲𝗮𝗿, 𝘄𝗲𝗹𝗹-𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱 𝘀𝗲𝗻𝘁𝗲𝗻𝗰𝗲𝘀 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗼𝗳 𝗔𝗜 𝘃𝗶𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆: use complete declarative statements, keep sentences under 20 words, add inline context, and use factual language over metaphors so both humans and models understand meaning quickly.

𝟯. 𝗟𝗟𝗠 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗰𝗮𝗻𝗻𝗼𝘁 𝗿𝗲𝗽𝗹𝗮𝗰𝗲 𝗦𝗘𝗢 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹𝘀: structure, depth, site performance, and technical foundations still determine visibility.

𝟰. 𝗛𝗲𝗮𝗱𝗶𝗻𝗴𝘀 𝗻𝗼𝘄 𝗮𝗰𝘁 𝗮𝘀 𝗮 𝗿𝗼𝗮𝗱𝗺𝗮𝗽 𝗳𝗼𝗿 𝗔𝗜: use descriptive main titles that state a full idea, break pages into question-driven subheadings, avoid vague labels, and maintain a logical broad-to-specific flow.

𝟱. 𝗟𝗲𝗮𝗱 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲 𝗮𝗻𝘀𝘄𝗲𝗿: start major sections with a short, definitive answer (30–80 words), then expand with detail - don’t bury the main point in explanation.

𝟲. 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝗹𝗶𝗻𝗸𝘀 𝘁𝗲𝗮𝗰𝗵 𝗔𝗜 𝗵𝗼𝘄 𝘁𝗼𝗽𝗶𝗰𝘀 𝗰𝗼𝗻𝗻𝗲𝗰𝘁: use descriptive link text, connect related themes, organize content around central pages with supporting articles, and avoid overwhelming pages with excessive links.

𝟳. 𝗠𝗮𝗸𝗲 𝗰𝗼𝗻𝘁𝗲𝗻𝘁 𝗲𝗮𝘀𝗶𝗲𝗿 𝗳𝗼𝗿 𝗔𝗜 𝘁𝗼 𝗲𝘅𝘁𝗿𝗮𝗰𝘁: clearly define key terms, use lists and numbered steps, add real FAQ sections, and clearly show that the page is regularly updated with visible “last updated” dates.

𝟴. 𝗦𝗽𝗲𝗮𝗸 𝘁𝗵𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗼𝗳 𝗔𝗜: use schema markup (machine-readable labels added to pages) to describe what the content is about, who published it, and what is being offered so models understand context, trust the source, and surface accurate answers.

𝟵. 𝗦𝗲𝗻𝗱 𝗰𝗹𝗲𝗮𝗿 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹 𝘀𝗶𝗴𝗻𝗮𝗹𝘀 𝘁𝗼 𝗔𝗜: use clean, accessible page structure and semantic markup (HTML tags that label page sections) along with natural-language titles, descriptions, and URLs so models can correctly interpret meaning and match content to real user questions.

𝟭𝟬. 𝗩𝗶𝘀𝘂𝗮𝗹𝘀 𝗶𝗻𝗳𝗹𝘂𝗲𝗻𝗰𝗲 𝗔𝗜 𝘃𝗶𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆: images and videos need captions, descriptive alt text (text descriptions for images), clear filenames, optimized metadata, and transcripts so AI can understand what they show and when to surface them.

Source: How to optimize your owned content for AI search – Linkedin

Subscribe now

Three very different strategies are emerging in agentic commerce. Each reflects a different bet on where power might sit when agents buy on our behalf.

𝟭. 𝗚𝗼𝗼𝗴𝗹𝗲 𝘄𝗮𝗻𝘁𝘀 𝘁𝗼 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝗶𝘀𝗲 𝘁𝗵𝗲 𝗲𝗻𝘁𝗶𝗿𝗲 𝗮𝗴𝗲𝗻𝘁 𝘀𝗵𝗼𝗽𝗽𝗶𝗻𝗴 𝗷𝗼𝘂𝗿𝗻𝗲𝘆:

• Google wants to define how AI agents discover products, compare options, access merchant information, and move into checkout.

• The ambition is to become the orchestration layer between AI assistants and merchants effectively shaping how buying decisions are structured before any money flow is triggered.

• If this works, Google will not only influence demand through search but also how transactions are initiated and routed.

• Tools: The Universal Commerce Protocol (a framework for how agents and merchants exchange structured commerce data), Gemini (Google’s AI assistant), and AI-enhanced Search.

𝟮. 𝗩𝗶𝘀𝗮 & 𝗠𝗮𝘀𝘁𝗲𝗿𝗰𝗮𝗿𝗱 𝘄𝗮𝗻𝘁 𝘁𝗼 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝗶𝘀𝗲 𝘁𝗵𝗲 𝗺𝗼𝗺𝗲𝗻𝘁 𝗼𝗳 𝗽𝗮𝘆𝗺𝗲𝗻𝘁:

• The card networks are taking a more focused approach: they are concentrating on what happens when an AI agent is authorised to spend money.

• Their strategy is to ensure that, even if AI changes how shopping happens, the execution of the transaction still runs on their rails.

• They are extending existing models of identity, authentication, and fraud protection to software agents but they are not trying to control discovery or product comparison.

• Tools: Visa Intelligent Commerce and Mastercard Agent Pay. These are new frameworks that allow AI agents to be securely authorised to make payments using existing card network infrastructure.

𝟯. 𝗦𝘁𝗿𝗶𝗽𝗲 𝘄𝗮𝗻𝘁𝘀 𝘁𝗼 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝗶𝘀𝗲 𝗵𝗼𝘄 𝗮𝗴𝗲𝗻𝘁𝘀 𝗲𝘅𝗲𝗰𝘂𝘁𝗲 𝗰𝗵𝗲𝗰𝗸𝗼𝘂𝘁:

• Stripe sees agentic commerce from an infrastructure perspective: when an agent decides to buy something, it needs a reliable, programmable way to complete the transaction.

• Rather than redefining the shopping journey or owning the payment network, Stripe focuses on making checkout easy for software to initiate.

• This positions Stripe as the execution engine regardless of whether discovery happens through Google, an assistant, or another platform.

• Tools: Stripe’s Checkout and payment APIs. These are widely used online payment tools that allow businesses (and now potentially AI agents) to trigger and manage payments through software.

Many see this as a race to build the best shopping assistant.

In reality, it’s a race to control the rails.

And we’ve seen this play before.

𝗦𝗼 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗶𝘀:

Who is best positioned to own the rails this time?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) 2025 FinTech Overview

What happened in Fintech in 2025 and what’s behind it?

Here is my behind-the-scenes summary based on the FT Partners 2025 Annual FinTech Almanac numbers.

𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗻𝗴 𝗮𝗰𝘁𝗶𝘃𝗶𝘁𝘆:

• Capital is concentrating into fewer, larger rounds as investors back proven platforms over early-stage fintechs.

• Profitability and predictable revenue now matter more than growth alone, as higher cost of capital has reset how risk is priced.

• Financial Management and WealthTech attract capital because banks and asset managers are still modernising core workflows around data, reporting, risk, and operations.

• Crypto funding has shifted from speculation toward infrastructure.

• Payments’ reduced share of financing reflects maturity of the core rails, with innovation moving to embedded and vertical-specific use cases.

• Banking and lending funding is spread across specialised tools (onboarding, underwriting, compliance, servicing, etc) as most banks choose to modernise in layers and not by replacing their core in one go.

• InsurTech investment is rising as insurers face worsening loss ratios (driven by climate volatility, inflation, and fraud) and use software to regain control over pricing, underwriting, and claims.

• Capital is increasingly flowing to markets that combine fast-moving regulation, public-sector capital, and national digital rails (real-time payments, digital ID, open finance).

• Mega-rounds are returning but mainly for scaled leaders, meaning this is not a generic market trend but focused on a small group of companies that already behave like infrastructure.

𝗠&𝗔 𝗮𝗰𝘁𝗶𝘃𝗶𝘁𝘆:

• M&A activity is accelerating because many fintech categories are now mature, making consolidation the fastest way to expand.

• Scaled fintechs are increasingly the buyers, using acquisitions to add capabilities faster than they could build internally.

• Acquisitions are focused on filling product gaps (risk, data, compliance, embedded payments, fraud) rather than buying growth.

• Payments M&A is driven by margin pressure and intense competition, with players buying scale and efficiency rather than chasing new geographies.

• Financial Management and WealthTech M&A is driven by demand for platforms that already sit at the centre of financial operations.

• Crypto M&A is selective, targeting regulated, compliant infrastructure rather than consumer-facing speculation.

• Cross-border M&A is rising as fintechs use acquisitions to enter regulated markets faster than licensing alone would allow.

• Private equity is accelerating as many strong fintechs generate cash but lack public-market scale, making them attractive candidates.

What are the trends that you see continuing in 2026? What did I miss?

Opinions: Panagiotis Kriaris, Graphic source: FT Partners

Subscribe now

3) Is Stripe worth $140 bn?

Is Stripe really worth $140 bn? Latest reports suggest Stripe is exploring a tender offer at a $140bn valuation.

𝗪𝗵𝗼 𝗶𝘀 𝗦𝘁𝗿𝗶𝗽𝗲:

• Stripe provides software and infrastructure for businesses to accept and process payments.

• It is one of the largest online payment processors globally.

𝗧𝗵𝗲 𝘀𝘂𝗰𝗰𝗲𝘀𝘀 𝗿𝗲𝗰𝗶𝗽𝗲:

• Stripe started with a clear, underserved problem: accepting online payments was slow, clunky, and overly technical, so Stripe built a simple, developer‑friendly API anyone could integrate in minutes.

• Chose developers as the primary audience, with great documentation, self-serve onboarding, and minimal code required to go live.

• Removed traditional pricing and contracting friction with transparent fees and easy setup, pulling in startups first and later enterprises.

• Expanded horizontally from core payments into subscriptions, marketplaces, billing, fraud‑prevention, and other financial building blocks, turning itself into economic infrastructure for the internet.

• Grew through strategic partnerships (Shopify, Amazon, big SaaS platforms, etc.) and fast global roll‑out, embedding itself into the backbone of modern digital commerce networks.

𝗩𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝗵𝗶𝘀𝘁𝗼𝗿𝘆:

• Mar 2021: $95 bn

Fuelled by COVID-driven online shopping surge and investor excitement for Stripe powering high-growth internet firms.

• Mar 2023: $50 bn

Rising interest rates deflated tech hype, forcing a valuation reset toward profitability and more realistic growth expectations.

• Feb 2024: $65 bn

Driven by improving margins, tighter cost discipline, and growing confidence in Stripe as durable financial infrastructure with a clear path to sustainable profitability.

• Feb 2025: $91.5 bn

Reflected continued margin expansion and investor confidence that Stripe’s growth is becoming more predictable, profitable, and resilient.

• Sep 2025: $107 bn

Driven by strong customer retention, expanding monetisation per customer, and confidence that new products (treasury, capital) deepen Stripe’s long-term moat.

𝗧𝗵𝗲 𝘁𝗲𝗻𝗱𝗲𝗿 𝗼𝗳𝗳𝗲𝗿:

A tender offer is a private transaction where existing shareholders (often employees and early investors) can sell some of their shares to new or existing investors.

𝗧𝗲𝗻𝗱𝗲𝗿 𝘃𝘀. 𝗜𝗣𝗢:

• An IPO would put Stripe into constant public pricing, heavy reporting requirements, and quarterly performance expectations.

• A tender lets Stripe give employees liquidity, avoid public-market volatility and lock-ups, and keep its focus on building product and platform.

• Stripe can control share price, buyer selection, and dilution, without raising capital given it is already profitable and cash-generative.

• Stripe gets liquidity and flexibility while keeping an IPO as a future option on its own terms.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Is an IPO off the table for 2026?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

4) BNPL Competition Map

BNPL is not a new offering. It is a new distribution model.

𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀:

• BNPL is unsecured consumer lending embedded into checkout.

• In most markets, BNPL is synonymous with instalments; in some markets, invoice-based BNPL remains dominant (e.g. Germany)

• The BNPL provider underwrites the transaction in real time

• The provider, not the merchant, takes the credit risk

• Revenue comes from merchant fees, interest on longer-term plans, and late fees

𝗧𝗵𝗲 𝗸𝗲𝘆 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲 vs traditional consumer credit is distribution:

• Traditional consumer credit is product-first: a credit card or a personal loan exists independently of the purchase and can be used across contexts.

• BNPL is context-first: credit is offered only at the moment of purchase, embedded directly into the checkout flow.

BNPL did not start as a platform strategy.

It started as a checkout optimisation and credit accessibility play.

𝗘𝗮𝗿𝗹𝘆 𝗽𝗶𝗼𝗻𝗲𝗲𝗿𝘀 𝗳𝗼𝗰𝘂𝘀𝗲𝗱 𝗼𝗻 𝗮 𝘀𝗶𝗺𝗽𝗹𝗲 𝗶𝗱𝗲𝗮:

𝟭. Offer short-term, interest-free instalments

𝟮. Remove friction from ecommerce checkout

𝟯. Serve consumers who either lacked access to traditional credit, or preferred alternatives

The first modern wave was led by branded, consumer-facing specialists such as Klarna, Afterpay, and Affirm.

These companies built recognisable consumer brands and positioned themselves as alternative ways to pay.

𝗧𝗵𝗲𝗶𝗿 𝗲𝗮𝗿𝗹𝘆 𝘀𝘂𝗰𝗰𝗲𝘀𝘀 𝗰𝗮𝗺𝗲 𝗳𝗿𝗼𝗺 𝘁𝗵𝗿𝗲𝗲 𝘁𝗵𝗶𝗻𝗴𝘀:

1. Tight checkout integrations

2. Fast, lightweight underwriting

3. Clear consumer value proposition (simple, transparent, often 0%)

As adoption grew BNPL moved from a niche ecommerce option to a mainstream checkout method.

What began as a narrow payment-and-credit feature became a strategic entry point into commerce and consumer finance.

𝗧𝗼𝗱𝗮𝘆 𝘁𝗵𝗲 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝘃𝗲 𝗹𝗮𝗻𝗱𝘀𝗰𝗮𝗽𝗲 𝗵𝗮𝘀 𝘀𝗵𝗶𝗳𝘁𝗲𝗱:

• Pure BNPL carries high merchant pricing, but economics are sensitive to credit losses, funding costs, and scale effects

• Rising interest rates exposed the capital-intensive nature of BNPL models and pressured valuations

• Regulatory scrutiny has increased as BNPL is increasingly treated as consumer credit, raising compliance costs

• BNPL is now widely available as a feature (Bigtech, PSPs, banks), reducing product-level differentiation

As a result, branded BNPL providers are diverging into distinct strategic paths:

1. Consumer commerce and shopping platforms

2. Credit-first consumer finance companies

3. Checkout distribution and merchant-scale plays

4. Regional consumer finance ecosystems

𝗧𝗵𝗲 𝗺𝗼𝘀𝘁 𝗶𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗽𝗮𝗿𝘁?

• Global BNPL e-commerce value grew ~252× in a decade: from $2.3bn in 2014 to $580bn in 2024!

• Despite this growth, BNPL still accounts for only ~5% of global e-commerce.

Opinions and graphics: Panagiotis Kriaris

Subscribe now

5) From AI to MAS

Everyone is talking about agentic AI and yet the next frontier is already in the making: Multi-Agent Systems (MAS).

AI didn’t arrive all at once – although in many cases it might seem it did. It evolved in distinct phases, each unlocking new capabilities and changing how work gets done:

𝟭. 𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗔𝗜 (𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝘃𝗲 𝗔𝗜):

• Systems powering rule-based models and statistical inference to detect fraud, recommend investments, and process documents - all in response to human prompts.

• Financial Services (FS) example: Credit scoring models and fraud detection engines improved efficiency, but remained passive tools waiting on human input.

𝟮. 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 (𝗚𝗲𝗻𝗔𝗜):

• LLMs and foundation models that brought language fluency and contextual understanding. These systems can create, explain, and summarize - moving from data crunching to content generation.

• FS example: Chatbots that summarize regulatory filings, generate client reports, or support advisors with contextual investment narratives.

𝟯. 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜:

• Systems that can interpret goals, plan actions, and operate independently within constraints. These agents shift the human role from executing tasks to defining intent.

• FS example: AI agents that autonomously rebalance portfolios based on client preferences and market movements - no human intervention required.

𝟰. 𝗠𝘂𝗹𝘁𝗶-𝗔𝗴𝗲𝗻𝘁 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 (𝗠𝗔𝗦):

• MAS represent the next leap. Multiple agents - each specialized - work together, negotiate, and adapt in real time to achieve shared outcomes across environments.

• FS: Agents handling client onboarding, AML checks, credit assessment, and regulatory filings collaborate seamlessly to approve new clients in minutes.

𝗪𝗵𝘆 𝘁𝗵𝗶𝘀 𝗺𝗮𝘁𝘁𝗲𝗿𝘀:

MAS enable distributed, intelligent systems that can self-organize, learn continuously, and respond dynamically to change. They reduce operational bottlenecks and shift digital architectures from static pipelines to adaptive, event-driven systems.

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗦𝗲𝗿𝘃𝗶𝗰𝗲𝘀:

• Efficiency: MAS collapse multi-day processes into seconds - from KYC to loan origination.

• Mass hyper-personalization: Real-time tailoring of product decisions across customer journeys and risk contexts.

• Resilience: Distributed agents can recover from local failures, reroute tasks, and maintain service continuity without manual intervention.

• Compliance: Agents track regulatory changes and trigger operational updates autonomously.

MAS aren’t just the next step in AI - they’re how AI starts to really work like a system. Value will increasingly come from turning multiple models into coordinated workflows that handle entire process and journeys end-to-end.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What do you think are the biggest challenges for MAS adoption?

Opinions: Panagiotis Kriaris, Graphic source: Capgemini

Subscribe now

Is AI eating the SaaS model?

Software stocks are down by double-digit percentages. And AI is to blame.

𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱:

• A broad market sell-off hit software, data and legal tech stocks globally after a wave of new AI tools from Anthropic sparked investor concern.

• The sell-off has wiped out hundreds of billions in market value - with Bloomberg estimating around $285 billion+ in losses across the sector.

• Major traditional legal and information services firms saw steep share price drops. Thomson Reuters - 18%, Pearson - 7%, LegalZoom - 20%.

𝗪𝗵𝘆:

• Investors interpreted the release of new AI capabilities as a signal that AI could automate work traditionally performed by licensed software platforms.

• Specific tools like the legal workflow automation plugin for Claude Cowork (Anthropic’s AI agent platform) were seen as threatening high-margin SaaS and data businesses.

• The fear isn’t just one tool - it’s that AI could erode recurring SaaS revenue models and replace certain workflow tooling altogether, prompting quick repricing of valuations.

𝗛𝗼𝘄 𝘁𝗵𝗶𝘀 𝗰𝗼𝘂𝗹𝗱 𝗽𝗹𝗮𝘆 𝗼𝘂𝘁:

• Model providers could move up the stack - from supplying intelligence to delivering complete business workflows.

• Foundation model companies increasingly control the reasoning layer, orchestration (agents), and user experience, which lets them assemble end-to-end enterprise tools without traditional SaaS vendors.

• Once a model provider owns the workflow, surrounding software risks becoming supporting infrastructure rather than the primary value layer.

• Investors are reacting to the idea that AI vendors no longer need software companies to reach enterprises - they can go direct.

• This doesn’t imply instant displacement. It implies gradual margin pressure and weaker pricing power for software categories with thin differentiation.

• If legal workflows can be built this way, then finance ops, HR, procurement, compliance, customer support, etc can follow the same pattern.

𝗧𝗵𝗲 𝗰𝗼𝘂𝗻𝘁𝗲𝗿-𝗮𝗿𝗴𝘂𝗺𝗲𝗻𝘁𝘀:

• Building real enterprise workflows is much harder than agent demos. And takes time.

• Enterprises don’t just buy capability. They buy reliability, liability, certifications, and support.

• Most workflows are deeply embedded in legacy stacks, making replacement slow and risky.

• Model providers still depend on software ecosystems for distribution and vertical depth - going fully direct everywhere is hard.

• In many categories, AI is a feature, not the product strengthening platforms more than replacing them.

• Historically, new platforms reshape value chains more than wipe them out.

It looks like a a complete reversal. Just a few years ago software was eating the world. Now AI might be eating software.

𝗜𝘀 𝗶𝘁 𝗮 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗼𝗳 𝗶𝗳 𝗼𝗿 𝘄𝗵𝗲𝗻?

Opinions: Panagiotis Kriaris, Graphic sources: Bloomberg, Bain

Subscribe now

2) Can PayPal reinvent itself?

PayPal’s CEO is out. But the question remains: can PayPal reinvent itself?

𝗟𝗮𝘁𝗲𝘀𝘁 𝗤𝟰’𝟮𝟱 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• Total Revenue: $8.68 billion, + 4% from last year but below the expected $8.78 billion

• Earnings Per Share (EPS) at $1.23, up slightly from last year but 5% below forecasts of $1.29

• Payment Volume: $475 billion for the quarter (up 9% year-over-year), but branded checkout (PayPal’s core service) grew only 4%

• Net Profit: $1.43 billion, up 28% but not enough to offset other results

• Venmo (PayPal’s peer-to-peer app): Processed $1.7 billion in revenue (up 20%), one of the few strong areas.

• 2026 Outlook: Profits per share flat, payment profits steady but not expanding, revenue low single-digits

𝗧𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 𝗯𝗲𝗵𝗶𝗻𝗱 𝘁𝗵𝗲 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• PayPal’s core PayPal-branded transactions grew just 4% in Q4 - persistent for 2+ years now, as merchants and consumers shift to faster alternatives like Apple Pay or one-click wallets.

• Revenue from powering others’ checkouts (e.g., Stripe-like services) faces fierce price competition; margins flat despite volume gains, a multi-year trend as Braintree volumes grow but profitability stalls.

• Active account growth trails rivals like Block (Cash App) or emerging wallets; PayPal added users but far slower than in its peak years.

• Apple Pay, Google Pay, BNPL and A2A rails fragment the checkout market - issues flagged in every earnings call since 2022.

• Cost cuts lifted profits per share, but sales growth stays stuck. Investors are tired of promises to grow profitably after years of missed targets and lowered forecasts.

𝗪𝗵𝗲𝗿𝗲 𝘁𝗵𝗲 𝗼𝘂𝘁𝗴𝗼𝗶𝗻𝗴 𝗖𝗘𝗢 𝗳𝗮𝗶𝗹𝗲𝗱:

• Failed to reignite branded checkout growth despite years of pledges to fight off Apple Pay and Stripe.

• Hyped innovations like Fastlane and Venmo upgrades, but they failed to deliver meaningful revenue jumps.

• Obsessed over short-term profit-per-share boosts via cost cuts, while overall sales stagnated.

• Oversaw an 80% stock drop since 2021 peak and lost investor trust through repeated forecast cuts.

𝗕𝗶𝗴𝗴𝗲𝘀𝘁 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝘂𝗽 𝗻𝗲𝘅𝘁:

• Shop apps like Amazon and Shopify build in their own one-click payments, now handling over 30% of online sales and bypassing PayPal.

• Instant bank transfers enable direct account payments without cards, reducing PayPal’s fees.

• BNPL options attract younger shoppers at checkout, while PayPal’s lags behind.

• AI shopping assistants negotiate prices directly with stores, skipping traditional checkouts like PayPal’s.

• Regulations on stablecoins and Apple Pay antitrust cases could force rivals to open up but also let them easily copy PayPal’s key strengths.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Can PayPal pull off a comeback or is it already a thing of the past?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) The AI-Native Bank Model

It is the biggest debate in banking today: how you integrate AI into the bank. Here is a summary of what works and what doesn’t.

𝗕𝗮𝗻𝗸𝘀 𝘄𝗵𝗲𝗿𝗲 𝗶𝘁 𝗱𝗼𝗲𝘀𝗻'𝘁 𝘄𝗼𝗿𝗸:

𝟭. AI initiatives emerge bottom-up from innovation or IT, with limited business ownership and unclear accountability for outcomes.

𝟮. Success is measured in pilots and demos rather than cost reduction, risk improvement, revenue, or customer impact.

𝟯. GenAI is treated as a silver bullet, driving focus on chatbots and copilots while deeper operational issues remain.

𝟰. AI is applied to isolated tasks instead of rethinking end-to-end processes, producing incremental gains only.

𝟱. Models are designed to inform humans rather than act in workflows, adding review steps instead of removing friction.

𝟲. Use cases are narrow and duplicated across teams, functions, and regions.

𝟳. AI sits outside core banking, payments, and risk platforms, limiting automation and integration.

𝟴. Data quality and ownership are not tested early, so problems emerge during deployment and are patched locally instead of being fixed at the process or source level.

𝟵. Governance and model risk are engaged late, slowing deployment and increasing friction.

𝟭𝟬. Scaling remains slow and costly due to tight coupling with local processes and systems.

𝗕𝗮𝗻𝗸𝘀 𝘄𝗵𝗲𝗿𝗲 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀:

𝟭. AI is tied to clear business priorities, with explicit assumptions on cost, risk, revenue, or service impact.

𝟮. Management owns AI outcomes and are accountable for delivery and value.

𝟯. Processes are simplified first, with AI introduced only where it materially improves outcomes.

𝟰. GenAI is used selectively and combined with rules, classical ML, and decisioning where predictability matters.

𝟱. AI outputs are embedded into workflows with defined human control points.

𝟲. Agent-based approaches are used selectively to coordinate complex processes, not as a universal model.

𝟳. Data improvements are driven by priority use cases, with ownership clarified where it matters.

𝟴. Risk and compliance are involved early to define controls and acceptable boundaries.

𝟵. Scaling happens by extending proven use cases to similar products or regions, reusing existing models, controls, and integration patterns rather than starting from scratch each time.

𝟭𝟬. Impact is measured against prior performance, allowing ROI to be quantified and compared to target KPIs.

𝟭𝟭. AI becomes operational in specific domains rather than a bank-wide transformation narrative.

Opinions: Panagiotis Kriaris, Graphic source: McKinsey & Company

Subscribe now

4) Fintech Q4’25 Top Equity Rounds

What do the latest top fintech equity deals tell us about the state of fintech?

𝗘𝗾𝘂𝗶𝘁𝘆 𝗱𝗲𝗮𝗹𝘀 𝗺𝗲𝗮𝗻:

• Investors are buying ownership in a company (vs. debt, credit lines, structured funding, etc).

• They back the business itself - not just its activity.

• They reflect belief in long-term value creation, not short-term volume or balance-sheet growth.

𝗛𝗲𝗿𝗲 𝗶𝘀 𝗺𝘆 𝘀𝘂𝗺𝗺𝗮𝗿𝘆 𝗼𝗳 𝘄𝗵𝗮𝘁’𝘀 𝗯𝗲𝗵𝗶𝗻𝗱:

𝟭. Fintech equity capital is concentrating into a small number of companies, with very large rounds replacing broad-based funding. This reflects a market where investors prefer to deepen exposure to perceived winners rather than diversify across many emerging players.

𝟮. The size of these equity rounds reflects a shift: capital is being used to reinforce long-term positioning, e.g. balance sheets, geographic expansion, regulatory capability etc than to fund near-term growth.

𝟯. Equity is flowing to fintechs embedded in core financial workflows, such as payments, spend management, payroll, wealth, and compliance. These businesses benefit from daily usage, high switching costs, and structural demand, making them easier to underwrite over long horizons.

𝟰. A clear separation is emerging between fintechs that earn money by processing or financing activity, and those that own platforms investors believe will shape how financial services operate long term. Equity capital is flowing almost exclusively to the latter.

𝟱. The prominence of payments and financial infrastructure indicates that investors continue to view these layers as foundational to the digital economy, even as margins compress and competition increases.

𝟲. Crypto-related equity funding is focused on infrastructure and payments-adjacent use cases rather than speculative or consumer-led products, reflecting a more mature focus on practical, real-world use cases across the sector.

𝟳. The mix of rounds suggests investors are funding companies to strengthen their position for the long run - whether that means being ready for public markets, staying private at scale, or becoming credible acquisition targets - rather than pushing for quick exits.

𝟴. Geographic concentration around the US remains strong, but the presence of globally scaled, cross-border platforms highlights continued demand for fintechs that can operate across markets and regulatory regimes.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Are the big-ticket rounds a sign of confidence or a sign of caution for Fintech?

Opinions: Panagiotis Kriaris, Graphic sources: CB Insights, Panagiotis Kriaris

Subscribe now

Can Open Banking replace cards? And what are the deciding factors?

To answer the question, we need to understand the mechanics behind each model.

𝗧𝗵𝗲 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝟰-𝗽𝗮𝗿𝘁𝘆 𝗺𝗼𝗱𝗲𝗹 (𝗩𝗶𝘀𝗮, 𝗠𝗮𝘀𝘁𝗲𝗿𝗰𝗮𝗿𝗱):

A typical transaction flow would look like this:

𝟭. A third-party (merchant) gateway identifies the right payment network

𝟮. The (card) network confirms the customer authorization

𝟯. The network routes the transaction to the customer’s bank (the issuer) for approval

𝟰. The network does the same with the merchant’s bank (the acquirer)

𝟱. After the approvals the gateway confirms the transaction

𝟲. The merchant’s bank settles with the network, whereas payment is done by the customer’s bank

For this middleman role, card networks define access fees (the costs for routing a card payment through their network) and interchange fees that merchants pay to the card issuers.

The latter can range from 0.3% for credit cards in Europe (there is a cap) and all the way up to 3.5% in the US and Canada (the most expensive in the world).

On the other hand, Open Banking can be a game changer because it connects merchants with consumers in a direct way, without intermediaries: payments are initiated directly from the consumer’s bank account.

𝗧𝗵𝗲 𝗢𝗽𝗲𝗻 𝗕𝗮𝗻𝗸𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹:

𝟭. The customer chooses ‘Pay by bank’ at checkout and selects the bank

𝟮. The transaction is redirected to the customer’s banking app

𝟯. The transaction is approved, and the merchant is notified

𝗪𝗵𝗮𝘁 𝘀𝗽𝗲𝗮𝗸𝘀 𝗶𝗻 𝗳𝗮𝘃𝗼𝘂𝗿 𝗼𝗳 𝗢𝗽𝗲𝗻 𝗕𝗮𝗻𝗸𝗶𝗻𝗴:

• PIS payments cost a fraction of cards, while delivering strong security, smoother UX, and near-instant settlement as real-time schemes scale globally

• Margin pressure is forcing change. As competition intensifies, merchants are actively looking for cheaper ways to accept and move money

• Open Banking is powering for the first time in a long period the build-up of a real, additional infrastructure layer that has the potential to influence the entire finance set-up.

𝗪𝗵𝗮𝘁 𝘀𝗽𝗲𝗮𝗸𝘀 𝗶𝗻 𝗳𝗮𝘃𝗼𝘂𝗿 𝗼𝗳 𝗰𝗮𝗿𝗱𝘀:

• Cards don’t just enable payments; they fund rewards, cashback, insurance, and loyalty ecosystems that actively shape consumer behaviour and are hard to sustain on low-margin A2A rails

• Decades of investment have built a dense card infrastructure - fraud management, disputes, tokenisation, wallets, and seamless checkout - that cannot be replicated easily

• Cards remain unmatched in global, cross-border, and offline acceptance, while open banking is still largely domestic

• Strong network effects and consumer habits keep cards embedded in everyday payments

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What are the lessons learned from Open Banking’s trajectory around the globe so far?

Opinions: Panagiotis Kriaris, Graphic source: Huntswood

Subscribe now

2) The UK Apple Pay Battle

This goes beyond digital wallets. The legal action against Apple in the UK is just the tip of the iceberg behind a complex structure spanning platforms, wallets, payments - and who is really in charge.

Apple has built a unique ecosystem:

• Apple Wallet is the built-in wallet on the iPhone, used to store payment cards.

• Apple Pay is the payment service that uses those cards to initiate contactless payments when an iPhone is tapped at a terminal.

• For that tap to work, the iPhone uses NFC, and Apple defines how wallets can access it through the operating system.

• Because Apple Pay is built into iOS, every iPhone tap-to-pay transaction is initiated inside Apple’s software before it reaches the bank or card network.

• Banks issue the cards, move the money and carry fraud risk, but must connect to Apple Pay to reach iPhone users.

• Only after Apple Pay has initiated the transaction does it travel to card networks and then to the merchant’s bank for processing and settlement.

What does this mean?

• This means Apple controls the starting point of the transaction, giving it influence over access, rules and economics - even though it never moves the money itself.

• The CMA has previously scrutinised Apple’s role as the iOS platform owner, including its control over NFC access for tap-to-pay.

• The issue is not Apple Pay as a consumer product, but Apple acting as a gatekeeper between users, wallets and the wider payments ecosystem.

What is the new legal case about?

• A new £1.5bn UK legal action challenges Apple’s role as the de facto gateway for iPhone contactless payments.

• It argues that banks have no commercially viable alternative but to accept Apple’s wallet terms if they want to reach iPhone users.

• As a result, fees and conditions set at the wallet layer are said to distort competition and spread costs across consumers, including those who do not use Apple Pay.

Apple’s position

• Apple says Apple Pay is optional and that banks are not forced to participate, arguing that participation reflects demand from iPhone users rather than lack of choice.

• It maintains that it does not charge consumers or merchants, and that any fees paid by banks reflect the value of enabling secure tap-to-pay transactions on iPhones.

• Apple argues that its control of tap-to-pay and NFC access is necessary to maintain security, privacy and system integrity on iOS.

• It points to recent changes allowing third-party wallets to access NFC and offer tap-to-pay functionality on iPhones, following regulatory scrutiny in the UK and EU.

This is a case that will not just test what happens with Apple, but how control at the wallet and device layer is treated across the wider platform economy.

What do you think happens next?

Opinions: Panagiotis Kriaris, Graphics: Codeburst, Panagiotis Kriaris

Subscribe now

3) Banks’ next IT transformation

It’s a paradox: banks have spent decades - and billions - trying to make their core systems intelligent. Now AI is moving intelligence out of those systems.

Banks spend around $650 billion every year on IT and digital transformation (McKinsey) - the size of Belgium’s GDP.

But how much of that spending has translated into measurable outcomes - ROI, growth, or better UX?

Not as much as the scale of investment suggests.

The challenge for banks is that this is no longer where transformation happens. AI is shifting value one layer above the core - in what used to be middleware, now evolving into an intelligence layer.

𝗪𝗵𝘆?

Because the core was never built for flexibility, but for reliability, control, and compliance. It’s designed to execute predefined rules, not to adapt or make new decisions - which is why intelligence is now moving one layer up.

𝗪𝗵𝗮𝘁 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝗺𝗲𝗮𝗻 𝗳𝗼𝗿 𝗯𝗮𝗻𝗸𝘀?

To capture this new value, banks must rethink transformation.
It’s no longer about replacing what works at the core, but about building intelligence on top of it.

In practice, this requires:

𝟭) Refocusing from core modernisation to intelligence enablement.

𝟮) Treating AI not as automation alone, but as a reasoning layer that connects strategy, operations, and experience.

𝟯) Embedding compliance, explainability, and control.

This is where a new breed of companies comes in - bridging the gap between static core infrastructure and real-time intelligence.

additiv is a prime example, having built an agnostic agentic and domain-specific intelligence layer - a platform that works with any core, adding reasoning, orchestration, and learning into banks’ existing frameworks with human oversight. It’s showing the way forward for operationalizing agentic AI in financial services.

𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀:

• System integration: Connects core, CRM, risk, and channel systems, keeping data consistent and product information aligned across lending, savings, insurance and wealth.

• Product delivery: Enables new propositions without touching the core - e.g., combining savings and investments or embedding credit at POS.

• Decisioning and personalisation: Uses customer data to recommend the next best product, adjust limits, or trigger advice.

• Compliance & governance: Auditability, explainability, and regulatory consistency across AI-driven outcomes.

• Innovation: New use cases - e.g. dynamic pricing, x-product bundles - can be deployed in weeks, reusing existing data and logic.

• Intelligent automation: applying domain rules, and semantically transformed data to orchestrate complex processes, e.g. E2E insurance claims.

The banks’ brains are moving outside the core. That’s only a threat if banks don’t react - for those that do, it’s an opportunity to gain flexibility and multiply the return on their IT investment.

Opinions: Panagiotis Kriaris, Graphic sources: additiv, Aperture

Subscribe now

4) The Capital One-Brex Deal

It’s the largest bank–fintech acquisition of all time. Capital One buying Brex is not just an M&A deal. Here’s what you won’t find in the headlines.

𝗪𝗵𝗼 𝗶𝘀 𝗖𝗮𝗽𝗶𝘁𝗮𝗹 𝗢𝗻𝗲:

• A major US consumer bank, built for the US scale and structure

• Its core strength is mass-market credit: issuing cards and loans at scale and managing risk with exceptional discipline

• Data-driven, highly standardised, compliance-first model

𝗪𝗵𝗼 𝗶𝘀 𝗕𝗿𝗲𝘅:

• A US fintech that runs a software platform to control and automate employee spending

• Built as a vertically integrated platform, combining corporate cards, spend management, and banking into a single system

𝗧𝗵𝗲 𝗱𝗲𝗮𝗹 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲:

• Capital One is acquiring Brex for $5.15bn

• Split in 50% cash and 50% Capital One shares

• $950m in deal, integration, and retention costs over the next three years

• Brex will be integrated into Capital One

• Closing is expected around mid-2026

𝗖𝗮𝗽𝗶𝘁𝗮𝗹 𝗢𝗻𝗲 𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗲:

• Gets access to a modern business payments and spend platform that would take years to build inside a regulated bank

• Shifts Capital One closer to daily business money flows (payments, spend, accounts)

• Adds software-led distribution and higher engagement with SMEs, beyond traditional lending

• Brex brings a proven operating model for product speed and iteration that complements Capital One’s risk and compliance strengths

𝗕𝗿𝗲𝘅 𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗲:

• Brex gains access to a large, stable balance sheet and a regulated banking infrastructure, instead of relying on partner banks

• Capital One provides distribution into the mainstream US SME market, beyond startups and tech companies

• Removes funding and regulatory constraints, letting Brex focus on product, platform, and workflow depth

• Being embedded inside a major bank gives Brex stability: longer time horizons, lower risk tolerance, and room to scale its model more broadly

𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• The deal adds to the evidence that build-vs-buy is no longer a real debate for banks

• Software and workflows are becoming more strategic than balance sheets in business banking

• Fintechs that sit in daily money flows are more valuable than those that only optimise financial products

• Regulation and scale still matter - but they now need to be paired with product speed and software depth

• The next phase of fintech consolidation will be about adding capabilities

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What is the biggest challenge that threatens the value of this deal?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

One of the clearest signals of market change is hiding in plain sight.

Private market valuations are diverging.

𝗧𝗵𝗲 𝗺𝗮𝗶𝗻 𝗺𝗲𝘀𝘀𝗮𝗴𝗲: not all growth is equal.

Here’s what the valuation shifts among some of the largest private companies reveal:

𝟭. The largest valuation gains are concentrated in a handful of companies, with the top four alone ranging from ~$350B to $500B. This is extreme capital concentration.

𝟮. All these incumbents share something in common: their global reach, capital-heavy models, and regulatory barriers drastically reduce the number of viable challengers.

𝟯. AI dominates valuation growth: OpenAI (+218% to ~$500B) and Anthropic (+1,802% to ~$350B) represent the sharpest valuation jumps.

𝟰. The scale of AI growth reflects investors pricing AI firms as future infrastructure layers with cross-industry reach.

𝟱. 𝗧𝗵𝗲𝗿𝗲 𝗶𝘀 𝗮 𝗵𝘂𝗴𝗲 𝗴𝗮𝗽 between AI and non-AI growth, which highlights a shift from funding many players to concentrating capital behind a handful of category leaders.

𝟲. There are 𝘁𝘄𝗼 𝘃𝗲𝗿𝘆 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗳𝘂𝗻𝗱𝗶𝗻𝗴 𝗽𝗮𝘁𝘁𝗲𝗿𝗻𝘀: a small group of companies seen as category owners attract disproportionate capital, while everyone else faces much tighter valuation ceilings.

𝟳. Investors are backing companies that control critical layers (compute, models, distribution) over marginal improvements in revenue or user growth, which is why capital increasingly concentrates in a small number of dominant players.

𝟴. Anthropic’s +1,802% increase illustrates how quickly valuations reprice when a company is seen as a credible alternative to the market leader.

𝟵. ByteDance (+60% to ~$480B) and SpaceX (+14% to ~$400B) show that at massive scale, even slower percentage growth translates into enormous absolute amounts.

𝟭𝟬. Scaled 𝗳𝗶𝗻𝘁𝗲𝗰𝗵𝘀 show 𝘁𝘄𝗼 𝗼𝗽𝗽𝗼𝘀𝗶𝗻𝗴 𝘁𝗿𝗲𝗻𝗱𝘀: 1) scale clearly improves revenues and efficiency, but 2) competitive and regulatory limits prevent those gains from translating into major valuation increases.

𝟭𝟭. Shein is a good example of how a scaled consumer platform can face constrained valuation due to external factors (regulation, supply-chain, geopolitics).

𝟭𝟮. Even globally scaled consumer software businesses like Canva struggle to generate large valuation gains, as additional scale does not materially change their competitive position or economics.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Is this the emerging AI divide - where AI-first players are (re)priced as future infrastructure, while everyone else faces capped valuation upside because they cannot meet the same expectations?

Opinions: Panagiotis Kriaris, Graphic source: CB Insights

Subscribe now

2) From search ads to chat ads

Sam Altman, Oct 2024: “I think of ads as a last resort for us as a business model.”

Jan 2026: ChatGPT launches ads.

THIS IS A BIG SHIFT.

𝗪𝗵𝗮𝘁 𝘄𝗮𝘀 𝗮𝗻𝗻𝗼𝘂𝗻𝗰𝗲𝗱:

• ChatGPT will display clearly labelled ads in the interface, outside the assistant’s responses

• Ads are contextual, not generated by the model and not allowed to influence answers

• Only free and low-cost plans will see ads; Plus, Pro, Business, and Enterprise remain ad-free

• Ads will be opt-out from personalization

• No ads shown in sensitive categories and no sale of chat data to advertisers

• Initial limited rollout, starting in the US

𝗪𝗵𝘆 𝘁𝗵𝗲𝘆 𝗵𝗮𝗱 𝗻𝗼𝘁 𝗱𝗼𝗻𝗲 𝗶𝘁 𝘀𝗼 𝗳𝗮𝗿:

• Ads risk undermining user trust by blurring the line between assistance and persuasion

• Commercial optimisation conflicts with user benefit: maximising advertiser value undermines neutral, high-quality outcomes

• Ad targeting pressures the use of conversational data (feels intrusive)

• Ads shift the product’s character from a neutral tool to a platform with commercial interests

• Premium positioning and subscription growth are harder to sustain with ads

• Early-stage focus was on model quality and safety, not monetisation optimisation

𝗪𝗵𝘆 𝗻𝗼𝘄?

• Running ChatGPT at scale is costly due to growing usage and more powerful models

• Subscriptions alone can’t fund unlimited free usage without strict limits or quality trade-offs

• Keeping ChatGPT widely accessible matters strategically, so ads became a way to subsidise non-paying users

• Users now expect AI to be cheap or bundled, reducing willingness to pay for a standalone tool

• The product is stable and trusted enough that ads can be added without fundamentally changing how answers work

• Ads can be shown around the chat rather than inside responses, making them less disruptive than before

𝗙𝗿𝗼𝗺 𝘀𝗲𝗮𝗿𝗰𝗵 𝘁𝗼 𝗰𝗵𝗮𝘁:

• Search ads are paid for access to attention, chat ads are paid for influence over decisions.

• In search, advertisers buy clicks and then convert elsewhere - in chat, advertisers pay to be present at the moment of choice

• Reduces the value of downstream traffic (websites, landing pages) and increases the value of the assistant itself

• Monetisation shifts from volume-based metrics (impressions, clicks) to outcome-based ones (selection, action, conversion)

• The assistant becomes the primary distribution channel

• OpenAI controls the chat interface, which means it controls access to customers and can capture more value, while other businesses become dependent on being surfaced inside ChatGPT

• For businesses, it raises the cost of not being integrated with AI assistants

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Can chat ads become the monetisation engine that reshapes an entire ecosystem?

Opinions: Panagiotis Kriaris, Graphic sources: Reuters, Google, OpenAI, P.Kriaris

Subscribe now

3) AI Agents in Finance

It’s one of the most radical changes AI agents bring to financial services - and yet it’s barely discussed.

Financial services have historically been built around systems, processes, and workflows:

• Products, pricing, and journeys were designed upfront with little room for variation.

• Decisioning relied on fixed rules and batch processing.

• Customer understanding was fragmented across systems with no coherent view.

As a result, the 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹 has always reflected system capabilities vs. customer needs:

• Broadcast-first: one message to many customers

• Limited channels used for different things: transaction alerts, compliance notifications, marketing campaigns, etc.

• Personalization limited to basic variables (name, balance threshold, segment).

𝗡𝗼𝘄 𝗔𝗜 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹𝗹𝘆 𝗰𝗵𝗮𝗻𝗴𝗲𝘀 𝘁𝗵𝗲 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝘀𝗲𝘁-𝘂𝗽:

• Real-time, contextual, continuously updated decisions

• Intelligence embedded directly into customer flows

• Systems that trigger and execute actions automatically

And that changes communication too, as it now needs to be part of the decision loop:

• Context-aware (driven by what happened)

• Intent-aware (based on the customer’s purpose)

• Adaptive (messages change based on response)

• Channel-native (tone, format, and interaction adapted per channel)

• Two-way by default

𝗧𝗵𝗶𝘀 𝗰𝗿𝗲𝗮𝘁𝗲𝘀 𝗮 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗮𝗹 𝗼𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝘆:

Move from broadcast messaging to agentic, intelligence-driven communication.

• Millions of customers

• Each receiving:

1) The right message

2) At the right time

3) On the right channel

4) With the right next action

Examples:

• Response-driven fraud warnings

• Payment reminders adjusted by timing and tone

• Adaptive onboarding flows

Players today need communication platforms that can execute across channels, at scale and in real time.

Sinch is one such example acting as the execution layer between internal systems and customer-facing interactions:

• Multi-channel message delivery based on system and agent instructions.

• High message volumes

• Two-way interactions, feeding customer responses back into automated flows

𝗪𝗵𝗮𝘁’𝘀 𝗻𝗲𝘅𝘁:

• One-way interactions are a thing of the past. Sinch data shows that nearly one in three consumers find them annoying.

• AI is already creating new types of communication

• Agent-to-agent communication is likely to be the next major shift

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What are the biggest showstoppers preventing mass communication from scaling?

Opinions: my own, Graphic sources: BCG, Panagiotis Kriaris

Subscribe now

4) 2026 Fintech IPOs

Which fintechs are most likely to go public in 2026? Here’s my list and the background.

1. Plaid

• Bank account connectivity for fintech apps

• IPO to position Plaid as essential financial infrastructure

• Reliance on banks and platforms it does not control

2. Stripe

• Global payments platform for online businesses

• IPO to provide funding for acquisitions

• Slower growth in core payments puts pressure on valuation

3. PicPay

• Brazilian consumer payments and digital banking platform

• The IPO gives it funding and credibility to compete with Nubank and

Mercado Pago

• Performance closely tied to Brazil’s economy

4. Revolut

• The closest we have in the west to a SuperApp

• Strong profitability has removed the main barrier to an IPO

• Managing regulatory and operational complexity across markets is a key risk

5. Kraken

• Crypto exchange serving both retail and institutional customers

• IPO to present Kraken as a regulated financial business vs. crypto

• Revenues remain cyclical with crypto markets

6. Monzo

• One of the largest consumer-focused digital banks in the UK

• IPO funds lending growth beyond private markets

• Competitive retail banking keeps margins under pressure

7. Airwallex

• Cross-border payments and treasury services for global businesses

• IPO to help build trust with large enterprise clients

• Strong competition from established global payments firms

8. Zilch

• BNPL and payments focused on regulated markets

• IPO to help distinguish Zilch from weaker competitors

• Regulatory scrutiny limits pricing flexibility and margins

09. PhonePe

• One of India’s largest digital payments platforms

• IPO monetises its scale and leadership position in daily transactions

• Highly dependent on regulatory and policy decisions in India

10. SumUp

• Payments and financial services provider for small merchants

• Public markets value its stable, recurring merchant revenues

• Core payments business is crowded and hard to differentiate

11. Starling Bank

• Profitable UK digital bank with strong SME and platform banking focus

• IPO to shifts valuation from fintechs to bank peers

• Growth rates are lower than consumer-focused challengers

12. Bitpanda

• Regulated European crypto investment platform

• IPO rewards its early investment in regulation and compliance

• Revenue remains tied to (volatile) crypto market conditions#

13. ConsenSys

• Software and infrastructure provider around the Ethereum ecosystem

• IPO to help value the company beyond crypto prices

• Business model is complex and not easy for public investors to assess

𝗪𝗵𝗶𝗰𝗵 𝗼𝗻𝗲𝘀 𝗱𝗼 𝘆𝗼𝘂 𝘀𝗲𝗲 𝗴𝗼𝗶𝗻𝗴 𝗽𝘂𝗯𝗹𝗶𝗰 𝗳𝗶𝗿𝘀𝘁?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

5) The Payment Gateway Shift

It is the most misunderstood layer in payments. Yet they are one of the biggest drivers of checkout conversion.

A payment gateway enables merchants to accept and process electronic payments securely.

𝗦𝗼𝗺𝗲 𝗺𝘆𝘁𝗵-𝗯𝘂𝘀𝘁𝗶𝗻𝗴:

• “Gateways are only about cards” → They orchestrate cards, wallets, BNPL, A2A

• “They are transaction pipes” → They manage authorization, routing, and recovery

• “Gateways all work the same” → The same transaction can succeed or fail depending on the gateway

• “They are backend plumbing” → They shape conversion and UX

𝗧𝗵𝗲 𝗽𝗵𝗮𝘀𝗲𝘀:

𝟭. Payment gateways were born in the early days of e-commerce to authorize card payments for simple, domestic online purchases:

• Gateway as a processor

• Enabled banks to approve online card transactions at scale

• Standardised merchant connections to acquiring banks

• Created the commercial foundation for online checkout

𝟮. As e-commerce went global, payment gateways enabled growth across markets and payment methods:

• Gateway as a payment platform

• Enabled international expansion without rebuilding payments

• Increased conversion through local and alternative payment methods

• Simplified operations across multiple providers and markets

𝟯. Now, as commerce moves inside software and vertical platforms, payment gateways are becoming decision-makers that directly shape revenue and margins:

• Gateway as a revenue optimisation layer

• Decide which provider, method, or route maximises acceptance

• Improve acceptance and reduce failed payments automatically

• Balance cost, risk, and conversion in real time

𝗧𝗵𝗲 𝘀𝗵𝗶𝗳𝘁:

Most payment companies (PSPs, acquirers, platforms, etc) don’t run their own payment gateway. Compliance, scale, and integration requirements make this expensive and complex.

So, they rent one.

These are known as 𝘄𝗵𝗶𝘁𝗲-𝗹𝗮𝗯𝗲𝗹 𝗴𝗮𝘁𝗲𝘄𝗮𝘆𝘀:

shared infrastructure that runs behind the scenes, branded and configured by the provider.

But not every white-label payment gateway is the same.

Some are just built to provide access, while more sophisticated ones are built to optimise success rates, directly impacting revenue.

Akurateco is a good example of the latter:

• Built as a white-label core, not a branded gateway wrapper

• Orchestration across multiple PSPs, acquirers, and payment methods

• Logic to route, retry, and recover transactions automatically

• Single payment core for multiple merchants, regions, and models

• Designed to improve approval rates at scale

• Capable of handling increasingly automated and agent-initiated transactions

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

In an AI-driven world, does owning the gateway logic matter more than owning the gateway brand?

Opinions: Panagiotis Kriaris, Graphic sources: Akurateco, Panagiotis Kriaris

Subscribe now

It is the biggest debate in US finance right now: should stablecoins be allowed to generate yield?

One side of the debate is led by banks and financial regulators.

𝗧𝗵𝗲 𝗰𝗼𝗻𝘁𝗿𝗮 𝗮𝗿𝗴𝘂𝗺𝗲𝗻𝘁𝘀:

𝟭. Paying yield changes a stablecoin from a payment instrument into a savings product, which has historically been regulated as a deposit

𝟮. This could trigger large-scale deposit migration. Bank of America’s CEO said up to $6 trillion could move from banks into stablecoins

𝟯. Yield-bearing stablecoins threaten core banking economics by pulling deposits away from banks’ low-cost funding base, forcing greater reliance on wholesale funding and reducing lending capacity

𝟰. Greater reliance on wholesale funding raises costs and volatility, weakening profitability and tightening credit availability

𝟱. In stress scenarios, stablecoin issuers lack central bank liquidity access and formal resolution mechanisms, increasing run risk

𝟲. Any systemic fallout would ultimately be borne by the public sector, while yield benefits accrue to private intermediaries

The other side of the debate is led by crypto firms, fintechs, and parts of the digital asset ecosystem.

𝗧𝗵𝗲 𝗰𝗼𝘂𝗻𝘁𝗲𝗿 𝗮𝗿𝗴𝘂𝗺𝗲𝗻𝘁𝘀:

𝟭. Yield alone does not make a stablecoin a deposit, because stablecoin reserves are not used to make loans or transformed into longer-term assets

𝟮. Stablecoins are fully backed instruments, not balance-sheet liabilities used for lending, so traditional deposit logic does not apply

𝟯. Preventing yield protects bank funding models rather than consumers, limiting competition from new forms of digital money

𝟰. Yield can be delivered through transparent, on-chain mechanisms or intermediaries without creating hidden leverage

𝟱. Restricting yield while allowing banks to pay interest creates an uneven regulatory playing field

𝟲. Regulation should focus on how stablecoin reserves are held and managed, rather than banning yield outright regardless of risk profile

𝗧𝗵𝗲 𝗽𝗼𝗹𝗶𝘁𝗶𝗰𝗮𝗹 𝗰𝗼𝗻𝘁𝗲𝘅𝘁:

• US lawmakers have been working on a broad crypto bill that, among others, touches directly on the stablecoin yield debate

• Draft language would prohibit crypto companies from paying interest simply for holding a stablecoin, while still allowing rewards tied to specific activities such as payments or loyalty programs

• The US Senate postponed progress after Brian Armstrong withdrew Coinbase’s support, arguing the provisions would effectively kill stablecoin rewards

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What do you think, should stablecoins be allowed to generate yield? And what the US senate will likely decide?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

2) The Apple Card 2.0 Deal

The saga is over. After a long process, Apple has chosen JPMorgan Chase as the new issuer of Apple Card. Here’s what’s behind.

𝗪𝗵𝗮𝘁 𝘄𝗮𝘀 𝗮𝗴𝗿𝗲𝗲𝗱:

• $20bn of Apple Card receivables move onto JPM’s balance sheet

• Portfolio transfers at a ~$1bn discount - rare for co-brands that usually sell at a premium - crystallising losses for Goldman

• JPM takes on all future Apple Card losses post-takeover, booking $2.2bn upfront for expected credit losses

• Goldman exits with a net positive earnings impact (~46 cents per share) as $2.48bn previously set aside for losses is released, more than covering the $2.2bn hit from selling the portfolio

• Mastercard remains the network

• Scope expands beyond credit into a broader Apple-Chase consumer finance bundle, including savings

• ~24-month migration window to move accounts, servicing, and risk infrastructure

• Marks the end of Goldman’s consumer credit push launched with Apple in 2019

𝗪𝗵𝘆 𝗝𝗣𝗠𝗼𝗿𝗴𝗮𝗻:

• JPM could absorb a $20bn+ Apple Card book with a lower-credit mix without forcing Apple to tighten eligibility or change product economics

• JPM accepted losses early (large upfront provisioning) instead of pushing risk back to Apple or demanding a prime-only reset

• Scale card issuers need industrial migration, servicing, and collections. JPM already runs this, while others would have had to build new capabilities

• Apple required a partner willing to operate fully behind Apple’s brand and product rules - not all issuers were comfortable with that

• American Express did not proceed due to the economics of a mass-market book and limited control over the customer relationship

• Synchrony lacked the balance-sheet depth and regulatory comfort for a BigTech-scale consumer platform

• Barclays faced strategic and regulatory limits in running a large US consumer credit program tied to Apple

𝗪𝗵𝘆 𝗠𝗮𝘀𝘁𝗲𝗿𝗰𝗮𝗿𝗱:

• $100m upfront economic incentive

• MC processes ~50% of Apple Pay volume, lowering the case for change

• Supported Apple Card’s original product design (no fees, Daily Cash, tight UX integration) without requiring product redesign

• Willingness to operate largely invisible to end users, letting Apple own customer relationship/UX

• Existing deep technical integration across Apple’s wallet, device-level security

• Tokenisation already tightly integrated, limiting value of complex network migration

• MC accepted underwriting/servicing decisions driven primarily by Apple

• Global acceptance/merchant coverage already optimal – no distribution upside

• Visa offered limited incremental value to justify migration costs

• American Express conflicted structurally with Apple’s model by combining network, issuer, and brand control

𝗧𝗵𝗲 𝗯𝗶𝗴 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Will / can JPMorgan succeed where Goldman failed?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) Google’s Agentic Play

The next play in the agentic commerce race is here. Google has launched the Universal Commerce Protocol (UCP). Here’s what it is and why it matters.

𝗧𝗵𝗲 𝗣𝗿𝗼𝗯𝗹𝗲𝗺:

• AI agents can recommend products, but can’t complete a purchase on their own

• Commerce APIs are fragmented, merchant-specific, and not agent-friendly

• Agents have to be built and maintained separately for each merchant

• Checkout, payments, and post-purchase flows are still built for humans

• Result: agentic commerce only works inside closed platforms (because they control everything)

𝗪𝗵𝗮𝘁 𝗨𝗖𝗣 𝗱𝗼𝗲𝘀:

• Gives merchants a standard way to describe their products, prices, and rules in a format AI agents can read

• Lets AI agents interact with merchants using the same structure, regardless of who the merchant is

• Allows an agent to create an order without custom code for each merchant

• Connects order creation to existing payment methods without redesigning checkout

• Allows agents to receive order status, updates, and post-purchase information automatically

𝗪𝗵𝘆 𝗚𝗼𝗼𝗴𝗹𝗲 𝗯𝘂𝗶𝗹𝘁 𝗶𝘁:

• People are starting to ask AI agents to shop for them instead of searching and clicking

• Today, Google can help users find products, but can’t help them complete purchases

• Without a shared format, scaling agent-driven shopping would require thousands of custom merchant integrations

• Google doesn’t control merchant checkout (merchants do), so it can’t solve this with a single app or service

• UCP lets Google influence how commerce works without becoming a merchant or payment provider

𝗧𝗵𝗲 𝗳𝗹𝗼𝘄 / 𝗵𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀:

1. A user asks an AI agent to buy something and sets basic preferences

2. The agent sends a UCP-formatted request to merchants describing the product, constraints, and preferences

3. Merchants reply using the same UCP structure, so offers are directly comparable

4. The agent evaluates the offers using consistent data instead of custom logic per merchant

5. The agent submits a UCP order request to the selected merchant

6. The merchant uses UCP to trigger checkout, payment, and order creation inside its own systems

7. Order confirmation and updates are returned to the agent via UCP

𝗜𝘀 𝗶𝘁 𝘁𝗵𝗲 𝗼𝗻𝗹𝘆 𝘀𝘂𝗰𝗵 𝗽𝗿𝗼𝘁𝗼𝗰𝗼𝗹?

• No, OpenAI has created the Agentic Commerce Protocol (ACP)

• UCP defines how agents and merchants exchange product information, rules, orders, and post-purchase updates

• ACP is mainly about letting an agent complete a purchase once a buying decision has already been made

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Who do you think is likely to launch the next protocol?

Opinions: Panagiotis Kriaris, Graphics: Google, Panagiotis Kriaris

Subscribe now

4) The Global Paymemts, Worldpay and FIS deal

The deal is now completed. It’s one of the most significant restructurings in payments in years. Here’s what you will not find in the headlines.

𝟭. 𝗗𝗲𝗮𝗹 𝘀𝘂𝗺𝗺𝗮𝗿𝘆

• Global Payments acquired Worldpay for ~$31B (net ~$22.7B):

- 45% from FIS for $6.6B

- 55% from GTCR for $24.3B

• FIS acquired 100% of Global Payments’ Issuer Solutions for $13.5B

• Announced in April 2025, the deal is officially completed as of 12 January 2026

𝟮. 𝗧𝗵𝗲 𝗽𝗹𝗮𝘆𝗲𝗿𝘀

• Global Payments Inc.: A top-3 global merchant acquirer and payment tech company

• FIS: A global technology provider in banking, issuer processing, and capital markets

• GTCR: A US private equity firm that acquired a majority stake in Worldpay in 2023

𝟯. 𝗥𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗲 𝗳𝗼𝗿 𝗚𝗹𝗼𝗯𝗮𝗹 𝗣𝗮𝘆𝗺𝗲𝗻𝘁𝘀

• Becomes a focused, pure-play merchant solutions player by exiting issuer processing

• Monetizes non-core business (issuing) that lacked scale and synergy

• Closes the e-commerce and enterprise gap, complementing its SMB and software-led model

• Gains scale at a discount - acquiring Worldpay at 8.5× EBITDA, well below peers

• Targets $600M in cost synergies and $200M in new revenue within 3 years

• Expands global reach in underpenetrated markets (Nordics, France, Japan)

• Improves product depth in card-not-present and cross-border

• Creates one of the largest global commerce platforms, processing trillions in annual volume across millions of merchant locations

𝟰. 𝗥𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗲 𝗳𝗼𝗿 𝗙𝗜𝗦

• Refocuses on core banking and issuing by exiting merchant acquiring - reversing the Worldpay bet and returning to its strength: serving financial institutions

• Adds credit processing scale, filling a key product gap

• Replaces a minority stake in Worldpay with a cash-generating, recurring revenue stream

• Unlocks $1.6B in cross-sell potential across top 150 global banks

• Strengthens its ability to bundle core banking, digital, risk, and card services

• Targets $150M in EBITDA synergies, with long-term upside

𝟱. 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗶𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀

• Focus on specialization: Global Payments focuses on merchants, FIS doubles down on financial institutions

• Full‑stack models lose appeal - scale and depth in a single domain now matter more than breadth

• Mid-tier players might face a strategic squeeze: without clear focus or global scale, many will struggle to compete

• M&A momentum: expect more carve-outs and targeted deals as firms streamline

• Embedded finance and issuer scale set the tone: whether you serve merchants or banks, delivering deeply integrated, high-volume capabilities is now the name of the game

• Competitive pressure on providers lacking either global scale or deep vertical specialisation will increase

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Is this a structural industry pivot from full-stack to specialization?

Opinions: Panagiotis Kriaris, Graphic sources: Global Payments, FIS

Subscribe now

5) The Future of Finance

This is big. It points to where financial infrastructure is heading - and how stablecoins are becoming core plumbing. And it goes far beyond Visa and BVNK. Here’s why.

𝗪𝗵𝗮𝘁 𝗶𝘀 𝗩𝗶𝘀𝗮 𝗗𝗶𝗿𝗲𝗰𝘁:

• Visa’s real-time push payments network, enabling payouts to cards, wallets, and accounts

• Runs on the same Visa card rails as traditional payments, but reverses the flow of funds: instead of a merchant pulling funds from consumers, they push money to recipients

• Used for payouts like gig economy earnings, refunds, insurance claims, remittances, and marketplace settlements

• It’s the largest real-time payments network of its kind, with reach across 190+ countries

𝗪𝗵𝗼 𝗶𝘀 𝗕𝗩𝗡𝗞:

• BVNK helps companies move money using stablecoins, without needing crypto expertise

• It takes care of the heavy lifting in the background - wallets, compliance, conversion, and settlement

• This lets businesses use stablecoins as a practical settlement and payout rail, not a crypto product

𝗧𝗵𝗲 𝗱𝗲𝗮𝗹:

• Visa Direct has enabled support for stablecoin-based payouts

• BVNK acts as the provider handling the stablecoin settlement layer

• Businesses using Visa Direct will be able to send payouts internationally, with funds settled in stablecoins rather than relying solely on traditional bank transfers

𝗪𝗵𝗮𝘁 𝗽𝗿𝗼𝗯𝗹𝗲𝗺 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝗱𝗲𝗮𝗹 𝘀𝗼𝗹𝘃𝗲?

• Visa Direct can deliver money instantly to recipients, but settlement between institutions still depends on bank operating hours, pre-funded accounts, and correspondent banking

• BVNK enables 24/7, real-time settlement using stablecoins, but does not itself control a global network for delivering funds to end recipients

• Global businesses need both: instant delivery to cards, wallets, or accounts and always-on settlement between institutions

• This deal combines Visa’s global delivery network with stablecoin-based settlement, without changing how businesses connect to Visa Direct

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• Stablecoins are no longer just an alternative rail - they’re becoming part of mainstream payment infrastructure moving moving from a crypto rail to a neutral layer for regulated payments

• Global payment networks like Visa Direct are starting to use stablecoins as a settlement tool, not a consumer-facing product

• It shifts the narrative from replacement (stablecoins vs. banks or cards) to enablement where stablecoins actually strengthen existing payment networks

• Over time, this will put pressure on correspondent banking, pre-funding structures, and traditional (cut-off based) settlement models

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

What do you think Mastercard will do in response?

Opinions: Panagiotis Kriaris, Graphic sources: Visa, BVNK

Subscribe now

1) Top 10 Fintech Trends in 2026

What are the trends that will shape fintech in 2026? Here is my top 10 list.

𝟭. 𝗦𝘁𝗮𝗯𝗹𝗲𝗰𝗼𝗶𝗻𝘀 as settlement infrastructure

Stablecoins will be used at the settlement layer of financial services to bypass cross-border delays, cut-off times, trapped liquidity, and fee opacity in existing rails. They will sit underneath banks and payment schemes for specific B2B, treasury, and platform pay-out flows rather than replacing them.

𝟮. Commerce and payments become increasingly 𝗮𝗴𝗲𝗻𝘁-𝗱𝗿𝗶𝘃𝗲𝗻

An increasing share of commerce and payment activity will be initiated by software agents outside of pilots, as shared protocols, governance models, and accountability frameworks compete for adoption across the value chain.

𝟯. 𝗙𝗶𝗻𝘁𝗲𝗰𝗵𝘀 reclaim the 𝗯𝗮𝗻𝗸𝗶𝗻𝗴 stack

More fintechs will pursue banking licences to gain direct control over deposits, settlement, and economics, primarily to reduce dependence on sponsor banks and external balance sheets rather than to operate as full-service banks.

𝟰. The rise of 𝗔𝗜-𝗻𝗮𝘁𝗶𝘃𝗲 fintechs

A new generation of fintechs is being built with AI embedded into core operations by default, allowing them to operate at lower marginal cost and handle higher volumes vs. legacy operating models.

𝟱. The 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝘄𝗮𝗿𝘀 escalate

As AI agents become widespread, incumbents and challengers will increasingly compete for control of the agentic layer. Incumbents will embed agents into existing platforms, while challengers will position agents above multiple services to capture distribution.

𝟲. Fintech play moves 𝗳𝗿𝗼𝗺 𝗯𝗿𝗲𝗮𝗱𝘁𝗵 𝘁𝗼 𝗱𝗲𝗽𝘁𝗵

Fintech competition will shift from broad coverage to execution within specific industries. Advantage will come from handling sector-specific cash flows, risk, and workflows, favouring embedded vertical players over horizontal platforms.

𝟳. Increased fintech 𝗰𝗼𝗻𝘀𝗼𝗹𝗶𝗱𝗮𝘁𝗶𝗼𝗻 activity

Fintech consolidation will increase as firms acquire capabilities rather than build them internally. Infrastructure providers will add vertical functionality, scaled technology firms will fill capability gaps, and incumbents will consolidate for defense.

𝟴. 𝗙𝗿𝗮𝘂𝗱 shifts to agent manipulation

Fraud will increasingly target agent-driven workflows rather than individual accounts or cards. Attackers will influence outcomes through input manipulation, synthetic interactions, and falsified context.

𝟵. Banks 𝘁𝗼𝗸𝗲𝗻𝗶𝘀𝗲𝗱 𝗱𝗲𝗽𝗼𝘀𝗶𝘁𝘀’ play

Banks will expand focus on tokenised deposits to retain control over settlement and liquidity, particularly in wholesale and treasury contexts.

𝟭𝟬. 𝗧𝗼𝗸𝗲𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻 moves to the infrastructure layer

Tokenization will advance where it improves core market infrastructure, with adoption concentrated in wholesale uses such as settlement, collateral management, and fund administration.

What’s missing?

Opinions: Panagiotis Kriaris

Subscribe now

2) Banks’ Profitability Playbook

It is the number one question among digital banks: how do you become profitable? Here’s my take on what actually works.

𝟭. A customer proposition that addresses core 𝗽𝗮𝗶𝗻 𝗽𝗼𝗶𝗻𝘁𝘀 of traditional banking, across both product and customer experience.

𝟮. 𝗠𝗮𝘀𝘀 𝗰𝘂𝘀𝘁𝗼𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻, leveraging social media and gamification.

𝟯. Low customer acquisition costs combined with high brand awareness, driven by 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝘃𝗲 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀 (digital-only, social-media-first, targeted campaigns, search and app-store optimization, rewards, etc.).

𝟰. Payment services are not enough. 𝗟𝗲𝗻𝗱𝗶𝗻𝗴 (interest income) is critical for profitability, which is why embedded approaches such as BNPL are especially attractive.

𝟱. The ability to generate 𝗳𝗲𝗲-𝗯𝗮𝘀𝗲𝗱 𝗶𝗻𝗰𝗼𝗺𝗲 by moving beyond the initial freemium model to a setup that charges transactional, account-based, FX, or penalty fees for value-added services, including subscriptions.

𝟲. A smart 𝗽𝗮𝗿𝘁𝗻𝗲𝗿 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝘆 that leverages the power of networks from a dual perspective: aggregation and referral (referral fees are a significant revenue driver).

𝟳. An 𝗲𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺 𝗽𝗹𝗮𝘆 that aims to build loyalty by creating a one-stop-shop, and subsequently monetizing the additional services (and revenue streams) that evolve around servicing customers on a lifestyle basis (beyond banking: transportation, shopping, dining, telecom, travel, etc.)

𝟴. A focus (and the associated know-how) on servicing dedicated 𝘃𝗲𝗿𝘁𝗶𝗰𝗮𝗹𝘀 (e.g. wealth management, SMEs), while expanding into adjacent and complementary offerings (e.g. insurance, investments, treasury).

𝟵. Ruthless 𝗰𝗼𝘀𝘁 𝗰𝗼𝗻𝘁𝗿𝗼𝗹 through operational leverage, automation, straight-through processing, and a technology stack designed to scale marginal costs towards zero.

𝟭𝟬. A clear 𝗳𝘂𝗻𝗱𝗶𝗻𝗴 𝗮𝗻𝗱 𝗯𝗮𝗹𝗮𝗻𝗰𝗲-𝘀𝗵𝗲𝗲𝘁 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝘆, particularly around deposits versus wholesale funding, cost of funds, and liquidity management.

𝟭𝟭. Risk management and 𝘂𝗻𝗱𝗲𝗿𝘄𝗿𝗶𝘁𝗶𝗻𝗴 𝗺𝗮𝘁𝘂𝗿𝗶𝘁𝘆, especially in credit, fraud, and AML.

𝟭𝟮. 𝗥𝗲𝗴𝘂𝗹𝗮𝘁𝗼𝗿𝘆 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝘆 as a competitive advantage, not a constraint: smart jurisdiction choice, passporting, licensing sequencing, and proactive supervisory engagement.

𝟭𝟯. 𝗖𝗼𝗺𝗺𝗲𝗿𝗰𝗶𝗮𝗹 𝗳𝗼𝗰𝘂𝘀, including clear customer segmentation and the willingness to exit products that fail to meet profitability thresholds.

𝟭𝟰. Organizational structure and 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 𝗱𝗶𝘀𝗰𝗶𝗽𝗹𝗶𝗻𝗲, with small, empowered teams, clear accountability, and fast decision-making.

Most digital banks don’t fail because they lack ideas.

They fail because they lack a proper profitability play.

Which ones have you seen work in practice?

Opinions: Panagiotis Kriaris, Graphic source: C-Innovation

Link to graphic source – a highly recommended analysis: https://www.c-innovation.eu/post/european-digital-banks-ranked-by-profitability-why-earnings-power-is-defining-digital-banking-leade

Subscribe now

3) Payment Credentials

Payments are moving from card numbers to digital access. This is a structural shift: transactions are decoupled from static card details.

Historically, the card number was the mechanism that made payments possible.

In payments jargon, it’s called the PAN (Primary Account Number) - the long number printed on a card:

• A single, static identifier

• Reused across all contexts: online, in-store, subscriptions

• Widely distributed across many intermediaries

• If stolen or exposed once, it can be reused anywhere

𝗧𝗵𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺:

The PAN was designed for a physical, card-present world, not for a digital environment of apps, wallets, and recurring payments.

The answer to this is 𝘁𝗼𝗸𝗲𝗻𝗶𝘇𝗮𝘁𝗶𝗼𝗻: instead of sharing the real card number, payments use a substitute value:

• The original card number is replaced with a token

• The token is used only in specific contexts

• The real card details stay hidden from most parties

• If a token is exposed, it is useless outside its intended scope

Despite its benefits, tokenization still has one big limitation: the payment still depends on the original card number - it’s protected, but not removed from the model.

𝗗𝗶𝗴𝗶𝘁𝗮𝗹 𝗰𝗿𝗲𝗱𝗲𝗻𝘁𝗶𝗮𝗹𝘀 remove the dependency on a reusable card number, allowing payments to be approved based on rules rather than shared details:

• No single identifier reused across all payments

• Credentials are created for a specific purpose

• Usage can be limited by time or context

• If compromised, the credential can be revoked or expire

𝗪𝗵𝗮𝘁 𝗰𝗵𝗮𝗻𝗴𝗲𝘀:

Transactions are no longer enabled by passing around a number but by issuing a digital permission for a specific use.

This is why the shift is structural: the card number is no longer the foundation of the transaction.

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

• As reusable numbers lose their central role, influence will shift toward issuers, wallets, and platforms.

• Cards won’t disappear, but they will stop being the reference model for digital payments. New payment flows will no longer need to copy card behaviour to fit existing infrastructure.

• When sensitive payment details are handled by fewer parties, merchants and intermediaries will no longer need to store them, and responsibility will move more toward issuers and networks that approve transactions.

• Payment innovation will be freed from the limits of what a card number can represent, enabling products built for subscriptions, embedded payments, and marketplaces.

• Existing payment rails will continue to move money, but the logic that governs how payments are initiated and constrained will shift upward, becoming the primary source of differentiation and competition.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

How are AI and agents influencing this shift?

Opinions: Panagiotis Kriaris, Graphic sources: Wikipedia, Panagiotis Kriaris

Subscribe now

Now that the year is over, it is the question to ask. What shaped fintech the most in 2025? Here’s my shortlist.

1. Instant Payments Regulation goes live in Europe

Real-time payments become the default, forcing banks and PSPs to upgrade infrastructure and controls.

2. HSBC closes its Zing app

Zing was HSBC’s attempt to compete with the likes of Revolut and Wise on FX for international payments

3. DeepSeek’s R1 AI model shocks the AI market

DeepSeek showed it can compete with OpenAI, challenging assumptions on compute, pricing, and AI market power.

4. FiDA, Europe’s Open Banking Framework, gets challenged

EU banks pushed back on FiDA over concerns it would give US Big Tech access to high-value financial data.

5. The war of credentials: Schemes vs. digital wallets

Visa’s Flexible Credential (VFC) and Mastercard’s One Credential (MOC) want to re-invent their card offering by consolidating multiple payment methods into one card.

6. The future of work is here

39% of current skill sets will be obsolete by 2030. 59% of workers will need reskilling.

7. Global Payments buys Worldpay and sells issuing to FIS

An industry shift away from full-stack models toward domain specialization as scale alone no longer guarantees advantage.

8. Revolut’s impressive 2024 results

Sustained profitability at global scale challenges the assumption that fintech growth and bank-like margins are mutually exclusive.

9. Agentic AI: PayPal vs Mastercard vs Visa

Control of payment credentials moves upstream into AI agents.

10. Stripe’s double bet

On AI as a control layer and on stablecoins as a programmable money layer.

11. Multi-Agent Systems (MAS) as the next AI evolutionary phase

Specialized multiple agents work together, negotiate, and adapt in real time.

12. GENIUS Act: Stablecoins go mainstream

Federal regulation integrates stablecoins into supervised financial infrastructure, reshaping competition and reinforcing US monetary influence.

13. JPMorgan monetizes data access

Pricing of customer-permissioned data challenges Open Banking economics and shifts bargaining power back to large banks.

14. EU banks launch Euro Stablecoin

Strategic response to USD stablecoin dominance aimed at preserving deposits, payment flows, and currency relevance.

15. OpenAI launches Instant Checkout in ChatGPT

Checkout now happens in the chat itself, not on a website.

16. Verification of Payee (VOP) becomes mandatory across SEPA

Instant payments shift fraud prevention at initiation, making payee verification a core control.

17. The agentic wars: Amazon vs. Perplexity

Amazon’s lawsuit against Perplexity signals a broader clash between closed platforms defending control and emerging AI agents challenging existing commerce rules

18. Banks’ crypto bet

Banks long skeptical on crypto are now betting on it to modernize their cash offerings.

What’s missing?

Opinions: Panagiotis Kriaris

Subscribe now

2) How to monetize a SuperApp

These days everyone wants to be a SuperApp. The very few that actually succeed have one thing in common: monetisation. Here’s how it’s done.

𝟭. 𝗔𝗻 𝗲𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺 𝗽𝗹𝗮𝘆 – not just access to multiple services – with a seamless, integrated, end-to-end experience across all aspects of modern life.

𝟮. 𝗣𝗮𝘆𝗺𝗲𝗻𝘁𝘀 as the undisputed underlying layer that connects everything else and acts as a base for the multitude of offerings on the platform.

𝟯. A wide range of 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗲𝗱 𝗽𝗮𝘆𝗺𝗲𝗻𝘁 𝗺𝗲𝘁𝗵𝗼𝗱𝘀 catering for different use cases and target audiences.

𝟰. Low 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗮𝗰𝗾𝘂𝗶𝘀𝗶𝘁𝗶𝗼𝗻 𝗰𝗼𝘀𝘁𝘀 as a direct result of the platform play, followed by aggressive up-selling and cross-selling of high-margin financial offerings and merchant added-value services.

𝟱. 𝗗𝗮𝘁𝗮 as the predominant tool for driving high engagement through context-aware, personalised offerings.

𝟲. A 𝘁𝘄𝗼-𝘀𝗶𝗱𝗲𝗱 𝗲𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺 (consumer + merchant) with the platform acting as the bridge that not only connects the two sides but fuels growth from one to the other in an open, two-way dynamic relationship.

𝟳. 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 and cloud services to a range of B2B partners (enterprises, telecoms, digital platforms, fintechs), turning the Super App into infrastructure.

𝟴. A 𝘀𝘂𝗯𝘀𝗰𝗿𝗶𝗽𝘁𝗶𝗼𝗻-led ecosystem for merchants: the platform becomes the enabling layer for partners, merchants and other tech providers to accept payments through a wide variety of instruments, creating predictable, recurring revenue and long-term lock-in.

𝟵. 𝗕𝘂𝗶𝗹𝘁-𝗶𝗻 𝗺𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝗰𝗵𝗮𝗻𝗻𝗲𝗹𝘀: merchants sell discount deals, gift vouchers and other digital goods like tickets to platform users.

𝟭𝟬. Leverage a 𝗻𝗲𝘁𝘄𝗼𝗿𝗸 𝗼𝗳 𝗯𝗮𝗻𝗸𝘀 and other FS providers to scale distribution faster than any single institution could on its own.

𝟭𝟭. 𝗙𝗶𝗿𝘀𝘁-𝗺𝗼𝘃𝗲𝗿 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 advantage with the local ecosystem. Paytm was, for example, the first app to launch UPI Lite in India and has subsequently enabled wallet interoperability that allowed full KYC Paytm Wallets to be universally acceptable on all UPI QR codes and online merchants.

𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻: which one of the above is hardest to execute in practice?

Opinions: Panagiotis Kriaris, Graphic source: paytm quarterly reports

Subscribe now

3) The AI Bank of the future

Becoming an AI bank is no longer a question of if but of how.

McKinsey’s framework is one of the most comprehensive on how to do it. Here are my main takeaways:

𝟭. Being an AI-first bank is about 𝟰 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹𝘀:

— Reimagining the customer experience (personalization, frictionless journeys)

— Using AI to augment decision making

— Modernizing core technology

— Setting up a platform operating model

𝟮. Banks excelling in AI 𝗱𝗼 𝟰 𝘁𝗵𝗶𝗻𝗴𝘀 𝘄𝗲𝗹𝗹:

— Set a bold, bank-wide vision for the value AI can create

— Transform entire domains, processes, and journeys vs. narrow use cases

— Build a full AI stack, increasingly powered by multiagent systems

— Sustain and scale value by setting up critical enablers of the AI transformation

𝟯. Capturing value from digital and AI requires a 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹 𝗿𝗲𝘄𝗶𝗿𝗶𝗻𝗴 𝗼𝗳 𝗵𝗼𝘄 𝗯𝗮𝗻𝗸𝘀 𝗼𝗽𝗲𝗿𝗮𝘁𝗲, across 𝟲 𝗰𝗼𝗿𝗲 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀:

— A business-led digital road map

— Talent with the right skills

— A fit-for-purpose operating model

— Technology that’s easy for teams to use

— Data that’s continually enriched and enterprise-wide

— Adoption and scaling

𝟰.𝗧𝗼 𝘂𝗻𝗹𝗼𝗰𝗸 𝗿𝗲𝗮𝗹 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘃𝗮𝗹𝘂𝗲 𝗳𝗿𝗼𝗺 𝗔𝗜, banks must:

— Choose the right scope of transformation by rewiring entire domains and subdomains

— Decide which subdomains to transform with AI and in which order

𝟱. 𝗕𝗮𝗻𝗸𝘀 𝗻𝗲𝗲𝗱 𝗮 𝗳𝘂𝗹𝗹 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝘀𝘁𝗮𝗰𝗸 - that goes beyond AI models - built 𝗮𝗰𝗿𝗼𝘀𝘀 𝟰 𝗰𝗼𝗿𝗲 𝗹𝗮𝘆𝗲𝗿𝘀, each requiring sustained investment to unlock real enterprise value:

— Engagement

— Decision making

— Data and core tech

— Operating model

𝟲. 𝗢𝗿𝗰𝗵𝗲𝘀𝘁𝗿𝗮𝘁𝗲𝗱 𝗺𝘂𝗹𝘁𝗶𝗮𝗴𝗲𝗻𝘁 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 are one of the most promising areas:

— Can automate complex decisions and workflows through a twofold use of AI

— Over time, banks could have hundreds of AI agents at their disposal, each trained to complete a particular task and ready to be called on by other agents or humans

— Example in preparing credit memos: productivity gains of 20%-60%, faster decision making by 30%

— Can form the basis of more engaging experiences for customers and bank employees

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Which one of the above will be most critical for banks in 2026?

Source: McKinsey & Company, Summary by: Panagiotis Kriaris

Link to the article: https://www.mckinsey.com/industries/financial-services/our-insights/extracting-value-from-ai-in-banking-rewiring-the-enterprise

Subscribe now

4) Core skills in 2030

The future of work is here. 39% of current skill sets will be obsolete by 2030; 59% of workers will need reskilling. Understanding the direction is critical.

This is my summary of the 2025 future of jobs report by the World Economic Forum:

𝗧𝗵𝗲 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• 22% of current jobs (∼ 262 million) will be affected.

• 170 million new jobs will be created and 92 million displaced, netting 78 million new roles.

𝗧𝗼𝗽 𝗗𝗿𝗶𝘃𝗲𝗿𝘀:

• Broadening digital access – 60% of employers expect it to reshape business by 2030.

• AI and tech advances – 86% cite AI as transformative; robotics (58%) and energy tech (41%) also key.

• Green transition – Climate action is driving demand for renewable energy and environmental roles.

• Demographic shifts – Aging boosts healthcare jobs; young populations fuel education demand.

𝗔𝗜’𝘀 𝗗𝗼𝘂𝗯𝗹𝗲 𝗜𝗺𝗽𝗮𝗰𝘁:

• While AI is expected to displace jobs, two-thirds of employers also plan to hire for AI-specific roles.

• Task automation will rise, but augmentation (human + machine collaboration) is seen as the more sustainable path.

𝗙𝗮𝘀𝘁𝗲𝘀𝘁 𝗚𝗿𝗼𝘄𝗶𝗻𝗴 𝗝𝗼𝗯𝘀:

• Big Data Specialists, AI/Machine Learning Specialists, FinTech Engineers, and Renewable Energy Engineers.

𝗙𝗮𝘀𝘁𝗲𝘀𝘁 𝗗𝗲𝗰𝗹𝗶𝗻𝗶𝗻𝗴 𝗝𝗼𝗯𝘀:

• Data Entry Clerks, Bank Tellers, Administrative Assistants, and Postal Service Clerks.

𝗖𝗼𝗿𝗲 𝗦𝗸𝗶𝗹𝗹𝘀 𝗼𝗳 𝘁𝗵𝗲 𝗙𝘂𝘁𝘂𝗿𝗲:

• Analytical thinking, resilience, adaptability, leadership, and AI/tech literacy.

• Curiosity and lifelong learning are rising in importance, with over 60% of employers identifying them as key skills for future roles—critical for adapting to evolving technologies and career pathways.

𝗪𝗼𝗿𝗸𝗳𝗼𝗿𝗰𝗲 𝗣𝗿𝗶𝗼𝗿𝗶𝘁𝗶𝗲𝘀:

• 85% of employers plan to upskill staff.

• Employee health and well-being is now a key talent attraction strategy.

Summary: Panagiotis Kriaris, Source: World Economic Forum

Subscribe now

It’s the big sequel. Banks vs Fintechs with AI at the center.

Is AI the lever for banks to finally close the innovation gap - or the catalyst for fintechs to scale?

𝗥𝗼𝘂𝗻𝗱 𝟭 was about disruption, fought primarily on the experience front:

• Fintechs reimagined the customer journey - faster onboarding, mobile-first UX, transparent pricing, unbundled products.

• Banks used scale and regulation as their structural moat - strong balance sheets, trust, and compliance capability, but slower, incremental modernization.

Now AI rewrites the rules of the game for 𝗥𝗼𝘂𝗻𝗱 𝟮:

• Shifting from UX and distribution to rebuilding the core architecture.

• Moving from staged product rollouts to continuous deployment cycles.

• Scale, once protecting margins through process and cost advantages, can now become an AI bottleneck.

McKinsey recently examined more than 600 AI initiatives launched between late 2022 and summer 2025 across both banks and fintechs. The analysis 𝘀𝘂𝗴𝗴𝗲𝘀𝘁𝘀 𝗮 𝘀𝗵𝗶𝗳𝘁 in how each side is approaching Round 2:

• Fintechs represent ~40% of companies in the dataset but account for nearly 70% of AI launches.

• Banks show far fewer AI launches, and many remain pilots rather than production deployments (e.g., internal service copilots, AML alert pilots).

• The fastest-growing launch clusters involve revenue-driving and agentic applications, led mainly by fintechs (e.g., automated credit decisioning, autonomous payment routing).

𝗪𝗵𝘆?

• Fintech economics require measurable revenue impact early, driving faster AI deployment. It’s AI as a growth driver vs. a cost lever.

• Banks’ core products already generate revenue at scale, so AI investments are scrutinized through risk and compliance lenses before value.

• Fintech product teams own end-to-end delivery, enabling rapid experimentation, while banks split ownership across risk/tech/business silos.

• Fintech architectures are modular, making it easier to plug in AI workflows, whereas banks’ stack dependencies mean each change touches multiple layers.

𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀:

For fintechs:

• AI can change fintech economics by allowing growth without hiring at the same pace.

• The challenge will be trust and oversight - fintechs must prove that their models are transparent enough to avoid regulatory and partner scrutiny.

For banks:

• Legacy operating models make it hard to move pilots into production, slowing meaningful AI impact.

• Scale and data will become advantages only if banks can restructure processes and responsibilities so AI can become embedded into core risk, credit, and compliance workflows.

𝗪𝗵𝗼 𝘄𝗶𝗹𝗹 𝘄𝗶𝗻 𝗥𝗼𝘂𝗻𝗱 𝟮?

Opinions: Panagiotis Kriaris, Graphic and data source: McKinsey & company

Subscribe now

It is one of the most widespread myths in banking: banks cannot play the ecosystem game.

I am not sure when I heard this for the first time, but it’s becoming increasingly hard to defend.

Not because banks have suddenly become good at building ecosystems (most haven’t), but because the reality on the ground - especially when you look at the breadth of tools and capabilities now readily available - tells a very different story.

𝗘𝗰𝗼𝘀𝘆𝘀𝘁𝗲𝗺𝘀 𝗮𝗿𝗲 𝗻𝗼 𝗹𝗼𝗻𝗴𝗲𝗿 𝗮 𝗰𝗵𝗼𝗶𝗰𝗲 𝘁𝗼 𝗺𝗮𝗸𝗲 𝗼𝗿 𝗻𝗼𝘁 – they have become the operating environment of the modern economy:

• Customers already live inside ecosystems (BigTech, platforms, Superapps). Banks are often just one service inside someone else’s flow.

• Value is created across journeys, not products - payments, lending, identity, data, all blend into a single experience.

• Distribution power has shifted. Whoever owns the interface and the data sets the rules - everyone else competes on margins.

• Standalone propositions are being eroded. Even strong offerings lose relevance if they are not embedded in a broader context.

This is exactly why 𝗼𝗽𝘁𝗶𝗻𝗴 𝗼𝘂𝘁 𝗶𝘀 𝗻𝗼𝘁 𝗮 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗳𝗼𝗿 𝗯𝗮𝗻𝗸𝘀.

• If you don’t own distribution, you become invisible infrastructure in someone else’s experience.

• Product excellence is no longer a guarantee for value - control of the customer journey has become more important.

• Even regulation is no longer the moat it used to be. In some cases, it accelerates disintermediation (think open banking access)

• Disintermediation doesn’t necessarily require disruption. You don’t need to be replaced to be sidelined - you just need to lose the primary relationship.

𝗔𝗹𝗹 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗹𝗲𝗮𝗱𝘀 𝗺𝗲 𝘁𝗼 𝘁𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝘀𝘁 𝗺𝗶𝘀𝗰𝗼𝗻𝗰𝗲𝗽𝘁𝗶𝗼𝗻 𝗼𝗳 𝗮𝗹𝗹:

playing the ecosystem game does not mean banks must become BigTech.

𝗕𝘂𝘁 𝗶𝘁 𝗱𝗼𝗲𝘀 𝗿𝗲𝗾𝘂𝗶𝗿𝗲 𝗮 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗺𝗶𝗻𝗱𝘀𝗲𝘁:

• Stop trying to own everything. Orchestrating third-party products can work extremely well.

• Banks don’t need to own the whole stack to win. They need to own a critical role in the value chain and be indispensable to the end outcome.

• Composable architecture is no longer nice to have - APIs, event-driven systems, and cores that plug directly into external journeys are a prerequisite.

• Choose your partners carefully. Most ecosystem value is captured (or lost) at the integration layer.

• Redefine success: from product P&Ls to distribution reach, embedded transaction volume, and frequency of use across external journeys.

So, 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗯𝗮𝗻𝗸𝘀 𝗶𝗻 𝟮𝟬𝟮𝟲 is not whether ecosystems make sense.

𝗕𝘂𝘁 𝗿𝗮𝘁𝗵𝗲𝗿:

Are banks willing to put in the effort to become part of the game – and do they really understand the cost of staying out?

Opinions: Panagiotis Kriaris, Graphic sources: Celent, Mambu

Subscribe now

This is the 2025 fintech unicorn list. But valuations, fundamentals, and investor expectations have completely changed. This is my take just before the year ends.

𝟭. 𝗩𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻𝘀 𝗵𝗮𝘃𝗲 𝗿𝗲𝘀𝗲𝘁

The 2021 highs were unsustainable. The correction that followed wiped out inflated multiples but forced companies to get serious about product-market fit, monetization, and real-world traction. A $5B valuation today means something fundamentally different than it did four years ago.

𝟮. 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗶𝘀 𝘁𝗵𝗲 𝗻𝗲𝘄 𝗻𝗮𝗺𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗴𝗮𝗺𝗲

The standout companies aren’t just building apps or enabling slick UX - They’re becoming essential layers in the financial services stack. APIs, orchestration layers, compliance tooling, embedded payments - these are the new growth drivers.

𝟯. 𝗧𝗵𝗲 𝗔𝗜 𝗽𝗹𝗮𝘆

Fintechs are no longer just experimenting with AI - they’re embedding it into risk, onboarding, personalization, fraud, and servicing. The most forward-looking unicorns are building agentic, decision-making layers that automate complexity at scale. It’s not about chatbots anymore - it’s about intelligent orchestration across the entire financial stack.

𝟰.𝗣𝗿𝗼𝗳𝗶𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗶𝘀 𝗸𝗲𝘆

The “growth at all costs” playbook is a thing of the past. Fintech unicorns are being pushed - by investors and the market - to show sustainable business models. Profitability (or a credible path to it) is now a baseline requirement.

𝟱.𝗧𝗵𝗲 𝗽𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝗽𝗹𝗮𝘆

Fintechs that started with a narrow focus - solving a specific pain point like onboarding, payouts, or FX - are now under pressure to broaden their scope. To reach and sustain unicorn status, they need to evolve into platforms: integrated, multi-service offerings that connect identity, payments, credit, and compliance. Scale now comes from depth, interoperability, and becoming indispensable across multiple touchpoints.

𝟲.𝗧𝗵𝗲 𝗨.𝗦. 𝗿𝗲𝗺𝗮𝗶𝗻𝘀 𝗺𝗮𝗿𝗸𝗲𝘁 𝗱𝗿𝗶𝘃𝗲𝗻

With no regulated Open Banking framework in place, U.S. fintechs scale by negotiating access rather than relying on standards. The model rewards speed and connections - but creates fragility. Moves like JPMorgan charging for data access show how quickly power can shift, making long-term innovation harder to sustain.

𝟳.𝗙𝗶𝗻𝘁𝗲𝗰𝗵𝘀 𝗹𝗲𝗮𝗱 𝗠&𝗔

Fintechs lead M&A probability rankings because they offer strategic utility at more grounded valuations. Their modular, API-driven models solve core problems banks and platforms would rather buy than build - making them ideal targets in a market shifting from hype to hard value.

Opinions: Panagiotis Kriaris, Graphic source: CB Insights

Subscribe now

It’s the big reversal. Banks long sceptical on crypto are now betting on it to modernise their cash offerings. JPMorgan is the latest. Why?

For years, large banks dismissed crypto as incompatible with core financial infrastructure. Their arguments:

• Extreme price volatility

• Regulatory uncertainty across jurisdictions

• Material compliance and custody risk

• Market structures dominated by speculation

• Repeated failures, hacks, and scandals

And yet, some of the biggest names are now doing exactly what they once argued against:

• 𝗝𝗣𝗠𝗼𝗿𝗴𝗮𝗻 just launched a tokenized money-market fund on a public blockchain, using Ethereum as the settlement layer.

• Earlier this year, 𝗚𝗼𝗹𝗱𝗺𝗮𝗻 𝗦𝗮𝗰𝗵𝘀, together with 𝗕𝗡𝗬, began tokenising representations of money-market fund shares on private blockchain infrastructure to streamline settlement and record-keeping.

So what does this actually mean?

𝗠𝗼𝗻𝗲𝘆-𝗺𝗮𝗿𝗸𝗲𝘁 𝗳𝘂𝗻𝗱𝘀 are among the safest and most conservative investment products in finance. They invest in short-term government debt and high-quality instruments, and are widely used by corporates and institutions as a cash-management tool.

𝗧𝗼𝗸𝗲𝗻𝗶𝘀𝗮𝘁𝗶𝗼𝗻 is the process of representing a real financial asset as a digital token on a blockchain. The asset itself doesn’t change. A money-market fund still holds the same short-term instruments - what changes is the plumbing.

𝗦𝗼 𝘄𝗵𝘆 𝗮𝗿𝗲 𝗯𝗮𝗻𝗸𝘀 𝗱𝗼𝗶𝗻𝗴 𝘁𝗵𝗶𝘀?

Not because they have changed their minds about crypto as an asset class.

They haven’t.

What has changed is the recognition that blockchain, used as infrastructure rather than as a market, can modernise how cash products are run.

𝗘𝘅𝗮𝗺𝗽𝗹𝗲, using JPMorgan’s announcement:

• In the past, money-market funds were run on traditional banking systems, with ownership updated at the end of the day and settlement taking time.

• JPMorgan is not changing the product, but how it is run: fund shares are now recorded as digital tokens on a blockchain.

• JPMorgan starts the fund with $100 million of its own money.

• This matters because investors are not going first - the bank has skin in the game from day one.

• Ethereum allows ownership and transfers to be updated in near real time.

• The result is faster settlement and simpler cash management, without changing the fund’s low-risk nature.

𝗧𝗵𝗮𝘁’𝘀 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝘀𝗵𝗶𝗳𝘁.

Banks’ systems are not built for the real-time economy. Blockchain can fix one of their biggest structural constraints: slow, batch-based settlement.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗶𝘀:

Is this a game smaller banks can realistically play – or it just widens the gap with big banks?

Opinions: Panagiotis Kriaris, Graphic sources: Reuters, WSJ

Subscribe now

2) Why is PayPal launching a bank?

PayPal has submitted an application to launch an industrial bank. But is it really a bank? And why does it make sense?

Just a few hours ago PayPal Holdings announced it has submitted applications to the Utah Department of Financial Institutions and the Federal Deposit Insurance Corporation to establish PayPal Bank, a proposed Utah-chartered industrial loan company.

𝗪𝗵𝗮𝘁 𝗶𝘀 𝗮𝗻 𝗶𝗻𝗱𝘂𝘀𝘁𝗿𝗶𝗮𝗹 𝗯𝗮𝗻𝗸?

• An industrial bank is a fully licensed bank

• It can take deposits and make loans

• Deposits are FDIC-insured, meaning: 1) customer deposits are protected (up to legal limits) 2) the bank is supervised for safety and soundness

• It is regulated by: 1) a state banking regulator, and 2) the FDIC (the US deposit insurance authority)

• From a customer perspective, it looks and behaves like a normal bank

𝗪𝗵𝘆 𝗨𝘁𝗮𝗵?

• Utah is one of the few US states that charters industrial banks, with a long regulatory track record

• It offers a clear, predictable framework that allows industrial banks to

operate nationwide

𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝗶𝗮𝗹 𝗯𝗮𝗻𝗸 𝘃𝘀. 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗯𝗮𝗻𝗸

• The bank itself is regulated like any other bank

• The key difference is ownership. An industrial bank can be owned by a non-bank company (like PayPal)

• This means the parent company: 1) does not become a traditional bank holding company 2) is not fully supervised by the Federal Reserve

𝗗𝗼𝗲𝘀 𝗶𝘁 𝗺𝗮𝗸𝗲 𝘀𝗲𝗻𝘀𝗲 𝗳𝗼𝗿 𝗣𝗮𝘆𝗣𝗮𝗹?

• Since 2013, PayPal has provided over $30bn in loans and working capital to more than 420,000 small businesses globally

• Small-business lending is already a core, scaled business for PayPal

• An industrial bank allows PayPal to deliver this lending more efficiently in the US by: 1) funding loans directly with FDIC-insured deposits 2) reducing reliance on third-party banks

• This lowers cost of funding, improves margins, and increases speed and control

• The proposed bank would also enable: 1) interest-bearing savings accounts 2) direct membership with card networks, reducing intermediaries in processing and settlement

PayPal wants to shift from being just a payments front end to owning more of the underlying financial infrastructure. The objective is greater control over critical layers of the stack that drive scale, resilience, and margins.

𝗧𝗵𝗲 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻:

Is a US bank charter becoming the next strategic play for scaled fintechs?

Opinions and graphics: Panagiotis Kriaris

Subscribe now

3) Can AI replace consultants?

If there is one industry ripe for disruption by AI, it’s consulting. But can AI really replace consultants?

AI is already doing (better) much of the work that junior consultants used to handle:

• Research - in hours vs. days

• Draft decks

• Faster market sizing, benchmarking, and scenario modelling

• Synthesis of large volumes of information

We have all received generic slides with recycled ideas and different logos. This is not a sustainable model anymore.

𝗔𝗻𝗱 𝘁𝗵𝗲 𝗽𝗿𝗲𝘀𝘀𝘂𝗿𝗲 𝗶𝘀 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗲𝘃𝗶𝗱𝗲𝗻𝘁:

• Accenture cut over 11,000 jobs in 2025 as part of a restructuring linked to AI reskilling and efficiency

• PwC reduced its global workforce by around 5,600 people in 2025 as consulting growth slowed

• McKinsey reportedly cut around 200 global technology roles in 2025, as more internal support work is automated with AI

AI is clearly changing the consulting business model.

That’s why the largest firms have moved aggressively to build dense networks of AI partnerships, investments, and acquisitions across the entire AI stack.

𝗕𝘂𝘁 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗻𝗼𝘁 𝗲𝗻𝗼𝘂𝗴𝗵.

Accenture has just announced a multi-year partnership with Anthropic, one of the leading developers of large language models, best known for Claude.

•~30,000 Accenture professionals to be trained on Claude models

• Accenture becomes a premier AI partner for Claude Code (Anthropic’s AI coding assistant for software development), making it available to tens of thousands of developers.

• Launch of joint offerings for CIOs to measure value and scale AI-powered software development

• Co-development of industry solutions, with an initial focus on regulated sectors (financial services, life sciences, healthcare, public sector)

𝗖𝗮𝗻 𝘁𝗵𝗶𝘀 𝗯𝗲 𝗮 𝗴𝘂𝗶𝗱𝗲 𝗳𝗼𝗿 𝗼𝘁𝗵𝗲𝗿𝘀?

Not every consulting firm can replicate Accenture’s scale, but the direction is clear:

• AI becomes embedded in delivery, changing project economics

• Large parts of junior execution work will disappear

• Consultants can no longer hide behind slides and frameworks

The old consulting model is breaking.

I don’t know what the future one will look like, but my experience suggests 3 things will stand out:

𝟭) 𝗧𝗿𝘂𝘀𝘁 - clients choose expertise and advise, and who delivers those will matter much more than in the past

𝟮) 𝗦𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 - deep industry, regulatory, or operational knowledge will gain value over broad, generic expertise

𝟯) 𝗦𝗸𝗶𝗹𝗹𝘀 - The profile of people that worked before won’t necessarily be a fit, as value will shift from producing work to making decisions

The real question is not whether AI will replace consultants, but which ones it will. What do you think?

Opinions: Panagiotis Kriaris, Graphic source: CB Insights

Subscribe now

4) Are we in an AI bubble?

It is (literally) the trillion-dollar question. Are we in an AI bubble? And is there an analogy to the dot-com bust?

𝗧𝗵𝗲𝘀𝗲 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗔𝗜 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• Nvidia reached ~$5 trillion market cap in 2025, the fastest value creation for a listed company in history.

•Nvidia now makes up ~8% of the S&P 500, the highest single-stock index weight since the 1970s.

• AI-linked megacaps generated ~70–80% of U.S. equity gains in 2025.

• Citi forecasts cumulative AI infrastructure capex by major tech firms above ~$2.8 trillion through 2029, with ~$1.4 trillion in the US alone.

• Roughly half of recent VC tech funding has flowed into AI startups, concentrated in a few companies.

𝗔𝗻𝗱 𝘁𝗵𝗲𝘀𝗲 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗱𝗼𝘁-𝗰𝗼𝗺 𝗻𝘂𝗺𝗯𝗲𝗿𝘀:

• Cisco’s market cap grew from ~$5 billion in 1993 to $555 billion by March 2000 - over 100x in seven years.

• At its peak in March 2000, Cisco made up ~3-4% of the S&P 500

• Tech stocks (internet/networking) drove most Nasdaq gains - the index rose ~400% from 1995 to March 2000 peak.

• Telecom/internet infrastructure capex hit ~$100–200 billion annually by late 1990s, fueling massive overbuild.

• VC funding saw ~40-50% of total dollars into internet/tech startups by 1999, concentrated in e-commerce and networking.

𝗧𝗵𝗲 𝗽𝗮𝗿𝗮𝗹𝗹𝗲𝗹𝘀:

• Rapid value creation concentrated in a single infrastructure provider (Cisco then, Nvidia now).

• Equity market gains increasingly dependent on a narrow group of megacap stocks tied to one technological narrative.

• Massive capex commitments ahead of proven monetization.

• VC and private capital flows clustering in one technology theme and consolidating around a small set of perceived winners.

• Infrastructure capacity being built faster than adoption can absorb.

• Investor expectations extrapolating current growth rates far into the future.

𝗔𝗻𝗱 𝘁𝗵𝗲 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀:

• AI has measurable productivity gains, unlike many dot-com models without revenue.

• Nvidia has hardware-software lock-in, where Cisco’s networking gear commoditized quickly.

• AI compute demand is concentrated in a few hyperscalers, not fragmented carriers.

• Chip shortages are keeping GPU prices high, whereas networking gear saw prices collapse quickly after 2000.

• AI adoption spans multiple industries, unlike dot-com’s narrow online commerce focus.

• Today’s leaders are profitable incumbents, not unproven startups.

• Consumer AI adoption scaled exponentially faster than early internet.

𝗦𝗼, 𝘄𝗵𝗮𝘁 𝗱𝗼 𝘆𝗼𝘂 𝘁𝗵𝗶𝗻𝗸 - is this an AI bubble, or are markets simply discounting future returns?

Opinions Panagiotis Kriaris, Graphic sources: WSJ, IMF, P. Kriaris

Subscribe now