# The Market Is Already Choosing for You
By Clint Sookermany | *The AI Value Institute*
---
This is the final edition of The Value Shift. Over the past 12 weeks, we have examined five structural forces reshaping financial services: agentic AI, open finance regulation, programmable money, portable digital identity, and the market restructuring they compound into. Each force is significant on its own. Together, they are permanently reclassifying who occupies which layer of the financial services value chain.
This restructuring was never in doubt. The evidence has been unambiguous since the first edition. What is still unsettled is narrower: most institutions have not explicitly chosen their position, and so the market is choosing it for them.
For most institutions, that default is already under way. Only 25 per cent of banks have integrated AI into their strategic playbook. The other 75 per cent are defaulting to a position they did not select. This edition explains where that default leads.
The industry is splitting into three archetypes. The middle ground is gone.
Every previous edition pointed toward this conclusion. Agentic AI is concentrating capability in the institutions with the largest data assets and technology budgets. Open finance regulation is mandating the data portability that makes AI agents genuinely useful. Programmable money is collapsing settlement from days to minutes. Portable digital identity is eliminating the onboarding friction that protected deposit bases for decades.
The combined effect is a permanent bifurcation. Three archetypes are emerging.
The first is Banking as a Platform. These institutions own the customer relationship, build AI-native data architectures, orchestrate across all payment rails, and issue digital identity credentials. Perhaps 10 to 15 institutions globally can credibly pursue this path. JPMorgan spent $17.1 billion on technology last year. Bank of America spent $13 billion. HSBC has deployed AI across more than 600 applications. The capital requirement alone excludes most of the world's 25,000 banks.
The second is Balance Sheet as a Service. The institution surrenders the customer interface to technology platforms and provides regulated capital, compliance infrastructure, and settlement. This is a viable model. It is also a commodity business. Margins are thin, counterparty risk is concentrated, and differentiation is limited to price and reliability.
The third is Niche Mastery. The institution dominates a specific vertical, geography, or capability where deep expertise creates a defensible position. This is the most underestimated archetype. It does not require platform-scale technology investment. It requires genuine specialisation and the discipline to say no to everything else.
The generalist bank that is neither platform nor utility nor specialist occupies the ground being destroyed fastest. The top five US banks already control 57 per cent of total banking assets and outspend regional competitors on technology by a factor of ten. That concentration will accelerate.
The platform banks are pulling away. The gap is now structural.
The evidence from Q1 2026 earnings is not subtle.
Six major US banks shed 15,000 employees while posting $47 billion in profits, an 18 per cent year-on-year increase. This is not cost-cutting. It is a structural shift in productivity driven by AI at operational scale.
JPMorgan now runs more than 500 AI use cases in production. Its LLM Suite reaches 250,000 employees, with approximately 125,000 active daily users. The system updates every eight weeks, feeding on more data from each business line. Its JADE data ecosystem provides the proprietary data advantage that commodity AI models cannot replicate. Goldman Sachs deploys Claude agents across $2.5 trillion in assets and reports three to four times developer productivity gains. Citigroup has 80 per cent of employees using AI tools, generating more than 42 million interactions.
Bank of America's Erica handles 98 per cent of customer inquiries without human intervention and drives 30 billion client interactions annually. Its "build once" architecture means the platform serves retail, corporate, and wealth management simultaneously. Digital sales reached two million transactions, representing 71 per cent of total sales.
These numbers describe a compounding advantage. More data trains better models. Better models improve customer experience and reduce costs. Improved experience and lower costs attract deposits and talent. Deposits and talent fund further investment. Regional banks cannot enter this cycle. The gap is no longer about capability. It is about the structural impossibility of catching up.
McKinsey estimates AI could generate $200 to $340 billion in annual value for global banking. BCG puts the figure at $370 billion by 2030. The top 200 global banks scaling AI could see 5 per cent revenue increases, 8 per cent operating cost reductions, and 16 per cent lower loan-loss provisions. A potential $289 billion combined benefit. The institutions that capture it are the ones already investing. The ones that are not are falling behind at a rate that compounds quarterly.
The Synapse collapse showed the fragility of undisciplined utility models
The April 2024 collapse of Synapse Financial Technologies is the cautionary case that every board discussing BSaaS should study.
Synapse operated as middleware between fintech applications and FDIC-insured partner banks, routing pooled customer funds through omnibus accounts. It served more than 100 fintech companies and reached 10 million retail customers. It was valued as a unicorn.
When it filed for bankruptcy, $265 million in customer funds were frozen. More than 100,000 customers were stranded. The bankruptcy trustee estimated a shortfall of $65 to $96 million between Synapse's records and bank records. The FDIC responded with a wave of consent orders against partner banks and proposed the "Synapse Rule" requiring beneficial-owner recordkeeping in custodial accounts.
The root cause was structural, not operational. BSaaS models that depend on middleware intermediaries create a three-body problem. The customer trusts the fintech. The fintech trusts the middleware. The middleware trusts the bank. No single entity has end-to-end visibility. When the middleware fails, the entire chain breaks.
This does not mean BSaaS is non-viable. It means BSaaS without direct regulatory oversight, rigorous ledger reconciliation, and genuine counterparty due diligence is a ticking clock. Post-Synapse regulation has made BSaaS more expensive, more constrained, and more concentrated among fewer well-capitalised institutions. The irony: increased regulation of BSaaS increases the concentration risk that regulators were trying to address.
73 per cent of fintechs fail within three years. The survivors became infrastructure.
The fintech industry has completed its Darwinian cycle. The numbers are conclusive.
Seventy-three per cent of venture-backed fintech startups fail within three years. The sector shed more than 60,000 jobs in 2025. Customer acquisition costs have become prohibitive against platforms with hundreds of millions of existing users and zero marginal acquisition cost.
The survivors share four characteristics. They sell to businesses, not individuals. They provide infrastructure, not applications. They have invested in bank-grade compliance. And their economics improve with every additional customer through network effects.
Stripe commands a $159 billion valuation and processed $1.9 trillion in payment volume in 2025, up 34 per cent year on year. Its Revenue and Finance Automation Suite is on track for $1 billion in annual recurring revenue by 2026. Adyen targets 20 per cent annual revenue growth with EBITDA margins above 55 per cent. Plaid connects 12,000 financial institutions and saw its valuation rise 31 per cent to $8 billion. These companies are the regulated plumbing of the new financial architecture.
The neobank story is more nuanced. Nubank serves 110 million customers with 32 per cent Latin American market share. Revolut has 30 million users. Monzo posted GBP 113.9 million in pre-tax profit. They have crossed the profitability threshold. But neobanks that remain single-feature face the same existential pressure as first-generation fintechs. They must evolve into platform players or be subsumed by tech giant embedded finance strategies.
M&A is accelerating the consolidation. More than 200 deals were announced in 2025. Global financial services deal values rose 25 per cent year on year. More than half of all fintech acquirers in 2025 were other fintechs, consolidating infrastructure to create full-stack platforms. Six US fintech IPOs raised $3.2 billion, the highest level in a decade, signalling that public markets are reopening for mature infrastructure players.
Compliance is the last durable moat
This was the central argument in our earlier analysis of the regulatory landscape, and it bears restating because it connects all five vectors.
The GENIUS Act, MiCA, eIDAS 2.0, PSD3/PSR, DORA, and FIDA have collectively raised the barrier to entry in financial services to historically unprecedented levels. Building AML/CFT controls, KYC infrastructure, data protection systems, and operational resilience frameworks to the standard these regulations demand requires investment that only well-capitalised institutions can sustain.
The result: compliance has become competitive differentiation. Institutions with strong compliance infrastructure can onboard new products faster, enter new jurisdictions sooner, and serve as trusted regulatory intermediaries for tech platforms that lack their own licences.
The GENIUS Act illustrates the dynamic. It creates a clear legal framework for stablecoin issuance but restricts it to regulated institutions. By classifying compliant stablecoins as neither securities nor commodities, it channels innovation through the banking regulatory perimeter. This favours institutions with both the licence and the technology to serve as issuers or custodians.
The embedded finance market is projected to grow from $126 to $146 billion in 2025 to $454 to $690 billion by 2030. Every dollar of that growth represents revenue migrating from traditional banking channels into platform ecosystems. The institutions that capture it will be those with the compliance infrastructure to underwrite it. The ones that cannot will watch the revenue leave.
If your board cannot answer these five questions, the market has already chosen for you
This is the decision framework that synthesises the entire programme. These five questions map to the five vectors we have examined. If your board cannot answer all five clearly, the strategic position is being decided by default.
One. Are you building governed agentic AI capabilities, or still running pilots? We documented that JPMorgan has more than 500 AI use cases in production. Goldman Sachs deploys autonomous agents across trillions in assets. The productivity gap between institutions at this scale and those running five pilot programmes is no longer a gap. It is a chasm that widens quarterly.
Two. Is your data architecture AI-ready? Institutions building to the EU's PSD3/PSR standard gain a single architecture that satisfies all three major jurisdictions. Those building to their home jurisdiction's minimum face compounding re-architecture costs. AI agents are only as useful as the data they can access. If your data architecture cannot feed them, your AI investment returns nothing.
Three. Can your infrastructure settle a transaction across fiat, stablecoins, tokenised deposits, and CBDCs? Cross-border settlement costs have collapsed from 2 to 7 per cent through correspondent banking to 0.1 to 0.5 per cent through stablecoin rails. Stablecoins processed $33 trillion in volume last year. That eclipsed Visa's $16.7 trillion. Treasury infrastructure that cannot speak stablecoin is already one generation behind.
Four. Are you positioned to issue identity credentials, or will you accept credentials issued by platforms? The eIDAS 2.0 acceptance mandate in December 2027 sets the clock. Apple Pay has 780 million projected users. Google Wallet has 520 million. The difference between issuing credentials and accepting them is the difference between owning the customer trust relationship and renting access to it.
Five. Have you explicitly chosen between Platform, Utility, or Niche? This is the question that subsumes the other four. If the answer is no, your institution occupies the middle ground. The evidence in every edition of this publication points to the same conclusion: the middle ground is where value is being destroyed fastest.
The window is 16 months. The decisions within it are irreversible.
The strategic window runs from now until approximately October 2027. Three factors make the decisions within it structurally irreversible.
AI-ready data architectures require three to five years to build. Institutions that have not started foundational work cannot complete it before competitive positions are locked in. Talent markets are tightening. Securing AI engineers and cross-domain compliance specialists will cost materially more in 2028. Customer relationships, once lost to platform intermediaries, do not return. When a customer's financial life is orchestrated by an AI agent within Apple Wallet or Google's ecosystem, the cost of recapture is orders of magnitude higher than the cost of retention.
JPMorgan's takeover of the Apple Card from Goldman Sachs (which lost approximately $1 billion on the partnership) demonstrates the dynamic. Even the world's largest bank will serve as a technology platform's backend when the alternative is losing the customer entirely. Meta has allocated $2.5 billion for stablecoin payments across WhatsApp, Facebook, and Instagram by Q4 2026. The platform economy is not coming. It is here.
Three questions for your board
1. Have you explicitly chosen Platform, Utility, or Niche? Is that choice reflected in your capital allocation, your technology roadmap, and your talent strategy? If the answer to any of these is no, the choice has not been made. The market is making it.
2. If you have not chosen, what evidence would need to be true for you to commit? This programme has documented the convergence of five structural forces across 170 cited sources and 50 named institutional case studies. If the evidence base is not sufficient, identify what is missing. If it is sufficient, identify what is preventing the decision.
3. What is the cost of another six months of strategic ambiguity? Quantify it. The platform banks are compounding their advantage at 30 to 40 per cent annually. The technology spend gap is widening. The talent market is tightening. The regulatory deadlines are approaching. Every quarter of indecision is a quarter of compounding disadvantage that cannot be recovered.
---
This is the final edition of The Value Shift. Over 12 weeks, we examined the five forces reshaping financial services: agentic AI, open finance, programmable money, portable identity, and market restructuring. The evidence is drawn from more than 170 cited sources and 50 named institutional case studies across The AI Value Institute's research programme on the future of banking and payments.
The forces are converging. The window is closing. The only remaining question is whether your board will act with the urgency the evidence demands.
If these questions are on your board's agenda, I would welcome a conversation.
Contact: hello@valueinstituteai.com | [valueinstituteai.com/contact](https://valueinstituteai.com/contact)
---
Clint Sookermany is Founder of The AI Value Institute (valueinstituteai.com). He advises financial services leadership teams on the convergence of AI, digital identity, and market restructuring.