The idea is the easy part. Whether it deserves your capital is the question.

Back to Insights

Every fintech starts the same way. Someone close to the industry watches a process fail, again, and thinks: I could fix that. The idea is the easy part. The expensive part is the decision that follows, whether to commit capital, your credibility and a year or two of your life to bringing it to market, and most founders make it on conviction when it should be made on evidence.

The market analysis chapter of my book exists for that decision. Its purpose is blunt: to find out whether you can make something customers want, that you can produce, and that will be profitable, before you spend as if the answer were yes. In 2026 the discipline matters more, not less, because the ground is moving. Gartner expects worldwide AI spending to reach $2.5 trillion this year, a distortion running through every market you might enter: it inflates buyer budgets, crowds your category with funded competitors, and changes what counts as defensible. Which makes the old questions sharper, and adds one the book could not ask in time.

Start with the problem, not the pitch

The chapter builds client personas across digital maturity, risk tolerance and regulatory exposure, and the exercise is routinely misused as a sales tool. Used properly it is a problem statement. The test of market discovery is whether you can say precisely whose problem you solve and what that problem costs them today, in a number. If you cannot name the number, you do not know the market yet, whatever the deck says.

The numbers exist, because the pain in financial services is measurable. Financial crime compliance costs institutions more than $206 billion a year globally, $85 billion of it in EMEA, and every CEO carrying a share of that cost wants it smaller. Fund managers are hunting alpha with new tools: 95 per cent already use generative AI and 58 per cent expect to increase its use in the investment process itself, with managers and allocators alike planning to spend more this year. And private markets are becoming liquid, a shift I wrote about in September: record evergreen fund launches and an ELTIF market growing at over 50 per cent a year, every new structure dragging behind it operational problems, monthly NAVs, retail-scale onboarding, liquidity management, that did not exist at this scale three years ago. Each of these is a persona done properly: a named class of buyer, a problem they already know they have, and a budget line the problem sits on.

Size it honestly

The chapter's TAM, SAM and SOM discipline is best understood as an honesty exercise. The worked example narrows from every asset owner on earth to European pension funds above $500 million with ESG mandates, to a realistic first target of five to ten early adopters in the UK and Nordics. TAM flatters; SOM decides. The pitch-deck habit of quoting the trillion-dollar addressable market tells an investor nothing except that you have not done the narrowing.

Proving demand means evidence someone else can touch. Ten named firms with the problem, the budget and a reason to move now. Design partners who give you their time before your product exists. A paid pilot, however small, because a buyer who pays a little has told you something no survey can. If you cannot assemble that list, you do not have a market yet; you have a hope, and capital committed to a hope is how eighteen months disappear.

The question the book could not ask

Porter's five forces frame the chapter's competitive analysis, and the threat of new entrants was always the uncomfortable one. It now has an entrant the book could not name: the model itself. JPMorgan has put its internal LLM platform in the hands of roughly 200,000 employees, and most large buyers are somewhere on the same road. So before you invest, ask the 2026 barrier question plainly: if this works, what stops an LLM recreating it in days, or a subject matter expert inside my target firm building it themselves with the tools their employer already pays for?

If the honest answer is nothing, the idea is a feature, not a company. This is the moat question, investors' shorthand for the durable advantage that survives a competitor knowing exactly how you work. What a model cannot conjure by Friday is specific: proprietary data nobody else holds, workflow depth that took years inside the regulated process to learn, regulatory permissions that take eighteen months to win, distribution into buyers who already trust you. That last one compounds slowest and holds longest, because in this industry the buyer is not just buying your product, they are putting their name behind it, and no model can generate that on demand. Your uniqueness is whatever survives this question, and if nothing survives it, better to learn that before the capital goes in.

One rode the wave. One drowned in the harbour.

The method is not academic; you can watch it separate winners from write-offs. Wise started with a problem its founders could cost to the pound: people moving money across borders were paying billions in FX spreads hidden inside "zero commission" transfers. That is a persona done properly, a named class of buyer and a number the problem cost them. The moat survived every version of the barrier question: licences across dozens of jurisdictions, direct connections into national payment systems from Brazil to Japan, and a price nobody with thinner infrastructure could match. By this year's results the network moved $243.5 billion across borders for 18.9 million customers while cutting its price again, to 0.52 per cent. No language model recreates that by Friday; the moat is made of regulatory permissions and plumbing, not code alone.

Bó is the other lesson. NatWest reportedly spent around £100 million building a standalone digital bank and closed it six months after launch with a customer base Monzo was adding in a fortnight. The failure was not execution; the app worked. It was market analysis skipped at the top: no problem solved that Monzo and Revolut had not already solved better, no differentiation beyond a parent brand the target customer did not want, and entry at the crowded end of the market with none of the forces at its back. Wise entered a market owned by banks and won on a problem they refused to solve; Bó entered a market owned by startups and lost because it brought no problem of its own. The chapter's discovery process would have approved one of these ventures and killed the other, and the killing would have been the more valuable decision.

Invest, or walk away

So the CEO's question, should I bring this to market, has a discipline. Invest when you can name the buyers, cost their problem, show early evidence they will pay, and point to a moat that survives the LLM question. Walk away, or reshape the idea, when the honest SOM is a handful of maybes, when the uniqueness is a wrapper, when the forces are against you. The frameworks are in the book; the discipline of using them before the money moves is yours. The market will answer almost any question you put to it properly. It is considerably cheaper to ask before you build.

Declan Sheehy is the founder of One Blackwater Consultancy and advises FCA-regulated fund and asset managers, and the fintech and RegTech firms that serve them, on GTM, AI strategy, governance, and operating models. He spent eighteen years scaling an alternative investments platform from $200 million to $31.6 billion. He is an IAPP-certified Artificial Intelligence Governance Professional (AIGP) and the author of Disruptive Innovation in Financial Services (2026).