Fintech Innovations Are Eating the Bank — And the Bank Is Helping

The moment a Vietnamese street vendor started accepting payments via QR code linked to a superapp wallet — bypassing every traditional bank in the country — was the moment I understood that fintech innovations weren't just disrupting finance. They were rewriting who gets to participate in it.

That was 2019. By April 2026, the rewrite is nearly complete in some markets, and only just beginning in others. The gap between those two realities is where the most consequential investment and policy decisions of the next decade will be made.

What's changed most sharply in the past 18 months isn't a single technology — it's the convergence of several threads that were developing separately: AI-native underwriting, real-time cross-border rails, embedded finance infrastructure, and the quiet but seismic shift in how central banks are treating digital currency. Together, they're producing a financial system that looks less like a modernized version of what we had and more like something architecturally different.

Here's what that actually means for investors, operators, and anyone trying to understand where the money is moving.

The Embedded Finance Layer Is Now Load-Bearing

If you've been following the embedded finance story, you know the pitch: financial services disappear into the platforms where people already spend time — e-commerce, logistics, HR software, ride-hailing. The bank becomes invisible infrastructure.

What's less discussed is how mature that infrastructure has become. Platforms like Stripe, Adyen, and their Asian equivalents (Razorpay in India, Xendit in Southeast Asia) have moved well beyond payment processing. They now offer lending, insurance issuance, treasury management, and FX — all via API, all white-labeled into non-financial apps.

The numbers reflect this. According to McKinsey's 2025 Global Payments Report, embedded finance revenues are projected to exceed $320 billion globally by 2028, up from roughly $65 billion in 2023. That's not a linear growth story — it's a compounding one, driven by the fact that each new embedded product increases platform stickiness and generates more data for the next product.

The strategic implication is important: the bank isn't being replaced, it's being pushed into the background. JPMorgan, DBS, and Standard Chartered aren't disappearing — they're becoming the licensed, regulated plumbing behind platforms that own the customer relationship. That's a massive margin and brand erosion problem for traditional banks, and a massive opportunity for anyone building the middleware layer.

I covered this dynamic in more depth in my earlier analysis of embedded finance — the core tension is that banks are simultaneously funding and being displaced by the fintech ecosystem they've chosen to partner with rather than fight.

AI-Native Underwriting: The Credit Revolution Nobody Is Talking About Loudly Enough

Here's a data point that should stop you mid-scroll: roughly 1.4 billion adults globally remain unbanked, according to the World Bank's Global Findex Database 2025. The traditional knock against serving this population was always risk — no credit history, no collateral, no reliable income documentation.

AI-native underwriting is dismantling that argument, and doing it faster than most regulators anticipated.

Companies like Tala (operating across Kenya, Philippines, Mexico, and India), Kredivo in Southeast Asia, and Lendio's AI-enhanced SME lending platform in the US are building credit models that bypass FICO scores entirely. Instead, they're using behavioral signals: app usage patterns, social graph data, transaction velocity, even the consistency of someone's GPS location relative to their stated employer.

The results are striking. Tala reportedly maintains default rates comparable to — and in some cohorts better than — traditional microfinance institutions, while approving loans in under three minutes. Kredivo's "buy now, pay later" infrastructure in Indonesia has extended credit to millions of first-time borrowers who would have been invisible to any conventional bank.

The fintech innovations driving this aren't magic. They're the application of large-scale behavioral modeling to populations that generate enormous amounts of digital data but have historically been excluded from the financial system because that data wasn't being read correctly.

The Regulatory Friction Is Real — And Necessary

This is where I want to push back on the pure optimism. AI underwriting at scale raises serious questions about algorithmic bias, data privacy, and the concentration of financial power in the hands of a few platform operators.

The EU's AI Act, which came into full enforcement in early 2026, classifies AI-based credit scoring as "high risk" — requiring explainability, human oversight, and regular bias audits. That's the right call. The problem is that most of the growth in AI underwriting is happening in markets where equivalent regulation either doesn't exist or isn't enforced.

When a fintech platform in Southeast Asia can make a lending decision based on your WhatsApp usage patterns without telling you, and without any regulatory body checking whether that model systematically disadvantages certain ethnic or geographic groups — that's not financial inclusion. That's financial surveillance dressed up in inclusion language.

Investors and operators need to be clear-eyed about this distinction. The platforms that build explainable, auditable AI credit models now will have a significant compliance advantage as regulation inevitably catches up. The ones that don't are building on borrowed time.

Real-Time Cross-Border Payments: The Infrastructure Shift That Changes Everything

SWIFT was built in 1973. The fact that it still forms the backbone of most international wire transfers — with settlement times of one to five business days and fees that can consume 5–10% of a remittance — is one of the most persistent inefficiencies in global finance.

That's finally changing, and the change is coming from multiple directions simultaneously.

Project Nexus, coordinated by the Bank for International Settlements, is linking domestic instant payment systems across countries — India's UPI, Singapore's PayNow, Malaysia's DuitNow, Thailand's PromptPay — into a single interoperable network. As of early 2026, the network covers roughly 1.7 billion people across participating economies. A migrant worker in Singapore can now send money to a family member in India in seconds, for a fraction of a percent in fees.

Meanwhile, stablecoin rails are emerging as a parallel infrastructure. Circle's USDC and Tether's USDT are already being used by remittance companies and SME importers in Latin America and Africa to move value across borders without touching the traditional correspondent banking system. The volumes are not trivial — on-chain stablecoin transfer volumes exceeded $10 trillion in 2024, according to Chainalysis data, though a significant portion of that represents automated DeFi activity rather than direct commercial transfers.

Central Bank Digital Currencies: The Wildcard

No discussion of cross-border payment infrastructure is complete without addressing CBDCs, and no discussion of CBDCs is complete without acknowledging how much uncertainty still surrounds them.

China's digital yuan (e-CNY) has processed over 7 trillion yuan in cumulative transactions as of late 2025, but adoption outside of government-mandated contexts has been slower than Beijing projected. The Bahamas' Sand Dollar, one of the world's first live retail CBDCs, has similarly struggled to displace cash and mobile money in everyday use.

The more interesting CBDC story appears to be in wholesale applications — central bank digital currencies used for interbank settlement and cross-border transactions between financial institutions, not retail consumers. The BIS's mBridge project, connecting CBDCs from China, Hong Kong, Thailand, and the UAE, has completed pilot transactions and represents what likely becomes the architecture for a parallel international settlement system that operates outside the dollar-dominated SWIFT network.

That geopolitical dimension is impossible to separate from the technology story. mBridge isn't just a payment efficiency project — it's infrastructure for de-dollarization, and the countries involved know it.

The Cloud Cost Problem Hiding Inside Fintech's AI Ambitions

There's a financial risk embedded in the fintech AI buildout that doesn't get enough attention: cloud spend is scaling faster than revenue for many AI-native financial platforms.

Every real-time credit decision, every fraud detection model running on transaction data, every personalized financial product recommendation — these are computationally expensive operations running continuously at scale. For a platform processing millions of transactions daily, the inference costs alone can be substantial.

This connects to a broader pattern I've been tracking: AI tools are generating cloud costs that organizations didn't budget for and often can't fully attribute. If you're building or investing in AI-native fintech, understanding the full cost architecture of your AI stack isn't optional — it's a survival question. This breakdown of how AI tools are generating cloud costs nobody budgeted for is worth reading carefully before you scale any AI-driven financial product.

The platforms that figure out how to run lean AI inference — through model distillation, edge computing, or smarter batching — will have a structural cost advantage over those that simply throw GPU compute at every problem.

What the Asia-Pacific Market Reveals About Where Fintech Goes Next

Having spent years covering Asia-Pacific markets, I'm convinced the region remains the best laboratory for understanding where fintech innovations land in practice versus theory.

Southeast Asia is particularly instructive. The region has some of the world's highest smartphone penetration rates alongside some of its most underbanked populations — a combination that creates the exact conditions where fintech can leapfrog traditional banking infrastructure entirely. Grab Financial, Sea's SeaMoney, and GoTo Financial in Indonesia have collectively built financial super-apps serving hundreds of millions of users who never had a conventional bank account.

India's UPI architecture — which processed over 18 billion transactions in a single month in late 2025 — demonstrates what happens when a government builds open, interoperable payment infrastructure and then steps back to let private innovation run on top of it. The result is a competitive ecosystem of fintech players building differentiated products on shared rails, rather than a winner-take-all platform war.

Korea presents a different case study. The "Korea Discount" narrative that has historically suppressed Korean financial stock valuations is slowly being challenged by a combination of governance reforms and genuine fintech innovation — Kakao Pay, Toss (Viva Republica), and KakaoBank have collectively demonstrated that Korean consumers will adopt digital financial services at scale when the UX is right. The question is whether the underlying governance and capital allocation reforms can keep pace with the product innovation.

The Trust Infrastructure Problem

Across all of these markets, the fintech innovations that scale successfully share one characteristic: they solve the trust problem before they solve the technology problem.

M-Pesa worked in Kenya not just because of mobile technology but because Safaricom's existing customer relationships provided a trust foundation that a new fintech startup couldn't have built from scratch. UPI worked in India because the government's backing gave users confidence that the system wouldn't disappear overnight.

This matters for anyone evaluating fintech investments or partnerships. Technology is table stakes. The durable competitive advantage in fintech is trust — regulatory trust, consumer trust, and increasingly, data trust. A platform that experiences a significant privacy or security failure doesn't just lose customers; it loses the foundational asset its entire business model depends on. The Fiverr privacy case is a useful reminder of how quickly that trust can collapse when a platform treats it as a secondary concern.

Three Actionable Takeaways

1. Watch the middleware, not the headline brands. The most durable value in fintech is accumulating in the infrastructure layer — the API providers, the compliance-as-a-service platforms, the real-time data pipes — not necessarily in the consumer-facing apps. Companies like Plaid, Stripe Treasury, and their Asian equivalents are building the connective tissue that every other fintech depends on.

2. Treat regulatory geography as a product variable. The EU's AI Act, India's data localization requirements, Singapore's MAS licensing framework — these aren't just compliance costs. They're market structure variables that determine which business models are viable in which markets. The fintech teams that build regulatory strategy into product architecture from day one will outperform those that treat it as a late-stage addition.

3. Demand explainability from your AI credit models now. Whether you're an investor doing due diligence or an operator building a lending product, the question "can you explain this credit decision to a regulator?" is coming. The platforms that can answer it clearly will have a significant advantage as enforcement tightens globally over the next 24 months.

The financial system being built right now — in the code of embedded finance APIs, in the behavioral models of AI underwriters, in the settlement protocols of cross-border CBDC networks — is not an incremental improvement on what came before. It's a different architecture, with different winners, different risks, and different questions about who gets access and who gets excluded.

The Vietnamese street vendor with the QR code isn't an edge case anymore. She's the template. The question is whether the system being built around her is one that genuinely serves her interests, or one that simply monetizes her data while calling it inclusion.

That distinction is worth fighting for — and it's where the most important fintech debates of the next decade will be won or lost.

NOCODE TECH STACKER