The first wave of fintech was about access: getting financial services to people and businesses that the incumbent financial system underserved or priced out. The second wave was about experience: rebuilding the interface layer of finance for the smartphone generation, making banking and payments feel like consumer software rather than institutional bureaucracy. We are now in the early stages of the third wave, and it is the most structurally significant of the three: embedding intelligence into the financial infrastructure layer itself, so that financial decisions are not made by humans with the assistance of financial data, but by systems that analyze financial patterns in real time and act autonomously.
To understand where this third wave is going, it helps to understand clearly where the second wave finished. Three companies, more than any others, define the high-water mark of fintech's second wave: Stripe, Brex, and Ramp. Each achieved multi-billion dollar valuations by solving fundamental problems in financial infrastructure for technology companies. And each is now pivoting toward the AI integration that will define their next chapter — while simultaneously creating the white space in which seed-stage AI-fintech companies are building the future they are not yet able to see from their position at scale.
Stripe: The API That Became the Financial Operating System
Stripe's founding insight, articulated by Patrick and John Collison when they started the company in 2010, was that accepting payments online should take hours, not months. The incumbent payment processors — First Data, Worldpay, and the card association direct programs — required weeks of underwriting, custom integration work, and ongoing compliance management that created a barrier to commerce that was both absurd and invisible to the incumbents who had grown accustomed to it. Stripe replaced months of friction with seven lines of code.
This sounds like a product improvement. What it actually was was a fundamental redistribution of economic value in the payments stack. By making payment acceptance trivially easy for developers, Stripe shifted the primary purchasing decision from the payment operations team to the engineering team, and from a procurement relationship to a developer experience evaluation. This was a category redefinition, not a feature improvement, and it is what produced Stripe's extraordinary growth trajectory.
By 2023, Stripe was processing over $1 trillion in annual payment volume, making it one of the most important financial infrastructure providers in the world by transaction throughput. The company's valuation, which peaked at approximately $95 billion in 2021 before a correction to the current estimated $65 billion range, reflects not just the payment processing volume but the breadth of the financial infrastructure stack Stripe has built over the past decade: billing, revenue recognition, tax calculation, fraud detection, banking-as-a-service, and most recently, an AI-powered suite of financial risk tools.
Stripe's AI investments are instructive for understanding where the AI-fintech convergence is most meaningful. The company's Radar fraud detection system, which uses machine learning trained on transaction patterns across millions of Stripe merchants, prevents an estimated $4+ billion in fraud annually — at a cost to merchants of a few basis points per transaction. This is a genuine AI use case that was impossible before neural network advances made real-time pattern matching at the transaction level tractable. The economic value is measurable, the performance improves with more data, and the data advantage Stripe has built through processing $1 trillion annually is essentially irreplicable by any startup.
This creates an interesting strategic picture for the seed-stage market. Stripe's AI advantage is strongest in exactly the domains where its data volume creates a moat: broad-pattern fraud detection, revenue optimization for large-scale merchants, and cross-merchant benchmarking. But it creates white space in domains where generalized pattern matching is less useful than specialized intelligence: vertical-specific credit underwriting, relationship-based SMB banking, and the application of AI to financial workflows that Stripe's horizontal infrastructure does not touch. These are the categories where we see seed-stage companies building with genuine competitive possibility.
Brex: Rethinking Credit Through the Lens of Technology Behavior
Brex launched in 2017 with an observation that the incumbent corporate card market had missed entirely: the creditworthiness of a technology startup is not well-predicted by the founders' personal credit scores or the company's operating history, because technology startups' primary assets — their venture funding, their ARR growth rate, their runway relative to burn — are not measured by traditional credit models. A Series A startup with $10 million in the bank, 200% year-over-year growth, and a burn rate of $400K per month was, by any reasonable analysis, an excellent credit risk. Traditional corporate card underwriting, anchored to historical cash flows and personal guarantees, said otherwise.
Brex replaced this with underwriting that read bank accounts, cap tables, and funding history. By treating technology-company financial behavior as its own category, requiring its own data model, Brex was able to offer corporate cards to companies that the incumbents declined — and to build a dataset of technology-company financial behavior that became increasingly predictive over time.
The company grew to a $12.3 billion valuation by 2022, serving tens of thousands of technology companies with corporate cards and a growing suite of financial management tools. The strategic pivot that followed — from serving all SMBs to focusing specifically on technology companies and larger enterprises — reflected a fundamental insight about where the data advantage was most defensible: in the specific behavioral patterns of technology-native companies, where Brex's training data and domain expertise gave it a genuine advantage that no incumbent or horizontal competitor could easily replicate.
Brex's AI roadmap focuses on using its accumulated data on technology-company spending patterns to make increasingly automated recommendations about expense management, travel policies, and vendor negotiations. The company has positioned itself explicitly as a financial intelligence platform for technology companies, not a card issuer with software features. This distinction matters for understanding where the AI-fintech convergence creates durable business models: the companies that win are those that use AI to provide genuinely differentiated financial intelligence, not those that add AI as a feature to an otherwise commoditized financial product.
Ramp: Intelligence as the Product
Ramp, founded in 2019, is the clearest example of a fintech company that built AI intelligence as a first-class product feature from the beginning, rather than retrofitting intelligence onto an existing financial product. The company's founding proposition was not "a better corporate card" or "a simpler expense management tool" but something more precise: "a system that actively helps companies spend less money."
This distinction is commercially significant. Corporate cards and expense management tools are fundamentally passive: they record what has been spent, categorize it, and help finance teams account for it. Ramp built a system that analyzes spending patterns and proactively identifies opportunities to reduce costs: duplicate SaaS subscriptions, vendor pricing that differs from benchmark rates, spending categories where negotiation could yield savings, expense policies that are generating friction without commensurate control benefit. The product is not a financial tool with AI features. The AI is the product — specifically, the intelligence that allows finance teams to operate more effectively with less manual analysis.
Ramp achieved a $5.8 billion valuation in 2023, becoming one of the fastest-growing fintech companies in history. Its growth was driven by a sales motion that would have been impossible without the AI core: Ramp could tell prospective customers, specifically and accurately, how much money their peer companies in the same industry and size bracket were spending on software vendors that Ramp customers had successfully renegotiated. The AI-generated benchmark data converted from a feature into a sales tool that created urgency and specificity in a category where both are typically absent.
For Swarm Capital, Ramp is the archetype of what AI-fintech convergence looks like at its most commercially effective: not AI applied to existing financial workflows, but AI that restructures the value proposition of the financial product itself. The companies building in this frame today are asking questions like: what financial decision currently requires significant human expertise and analysis, and how can we make that decision available to every company regardless of the sophistication of their internal finance team?
The Whitespace: Where the Third Wave Is Building
Stripe, Brex, and Ramp collectively built the financial operating system for technology companies in the 2010s and early 2020s. Their success created the infrastructure on which the third wave is building — and their scale created blind spots that seed-stage companies can exploit. Understanding those blind spots is central to how Swarm Capital evaluates fintech opportunities in 2026.
The most significant structural gap in the current fintech landscape is vertical-specific financial intelligence. Stripe, Brex, and Ramp are horizontal platforms: they serve technology companies broadly, which means their AI models are trained on the average of technology-company behavior. But financial intelligence that matters for a healthcare company looks very different from intelligence that matters for a construction company or a logistics company. Insurance reimbursement patterns, job cost accounting, freight billing complexity — these are domains where specialized intelligence outperforms general intelligence, and where the data moats are accessible to early-stage companies because the incumbents have not yet invested in building them.
We are also seeing significant opportunity in the application of large language models to financial document processing. The financial system generates enormous volumes of unstructured documents — contracts, invoices, earnings calls, regulatory filings, loan applications — that contain financially material information that is not currently being systematically analyzed by AI. The companies building language model pipelines specifically designed for financial document categories — with fine-tuning on domain-specific data and structured output formats designed for financial decision systems — are building competitive moats that will be extremely difficult to displace once they accumulate sufficient labeled data.
The third category we track closely is AI-embedded compliance and regulatory technology. Financial regulation is both extensive and increasingly complex, particularly in areas like anti-money laundering, know-your-customer verification, and cross-border payment compliance. The compliance function in financial services institutions consumes an estimated 10-15% of total operating costs — and the majority of that cost is human review of cases that an AI system could resolve with comparable or superior accuracy. The companies building AI systems specifically designed to handle regulatory compliance workflows are addressing a cost center that financial institutions are highly motivated to reduce, and building in a category where specialized domain knowledge creates barriers to entry that general-purpose AI providers cannot easily cross.
Seed-Stage Signals in AI-Fintech
At Swarm Capital, our evaluation framework for AI-fintech seed investments focuses on three questions that the Stripe, Brex, and Ramp case studies illuminate.
First: is the AI a feature or the product? Companies that use AI to improve an existing financial product (slightly better fraud detection, marginally faster KYC) are competing on a dimension where scale advantages favor the incumbents. Companies where the AI capability is the reason customers switch — where the value proposition literally does not exist without the intelligence layer — are building in territory where seed-stage companies can establish genuine differentiation.
Second: does the data flywheel reinforce the competitive position? The most defensible AI-fintech companies improve their models with every transaction, document, or decision they process. This creates a compounding data advantage that makes the product objectively better at incumbent-relevant tasks for each customer processed. Brex's underwriting model improved with every technology-company credit decision. Ramp's benchmark data became more accurate with every vendor negotiation completed by a Ramp customer. The companies we back demonstrate this flywheel in their first year of operation, at volumes that make the improving performance statistically significant.
Third: is the go-to-market motion aligned with the AI value proposition? The most common failure mode in AI-fintech is building a technically impressive AI system and then selling it with the same motion as a traditional SaaS product. Ramp's success was partly attributable to its go-to-market innovation: using the AI-generated benchmark data as a sales tool that created immediate, personalized urgency. The seed-stage companies with the best AI-fintech outcomes are those where the founder understands not just how to build the intelligence but how to make the intelligence legible to a CFO or financial controller who needs to justify a purchasing decision to a board.
The AI-fintech convergence is not a trend that will play out over decades. The infrastructure is largely in place, the talent pool has expanded dramatically, and the incumbents are both too large to move quickly and too aware of the threat to ignore it. The window for seed-stage companies to build the specific, data-rich, vertically intelligent fintech platforms that address the gaps left by horizontal incumbents is open now — and the companies that establish data flywheels and customer workflows in the next eighteen to thirty-six months will be extremely difficult to displace from the market positions they secure. That is the investment opportunity Swarm Capital is actively pursuing.
