When Software is Built On Demand, What is a Fintech for?

By Koen Vanderhoydonk, CEO at The Connector.

This week, a head of innovation at a bank told me that fintech vendors might not be needed for much longer. His logic was straightforward. If the bank can describe what it needs and have an application built within days, why buy one from someone else? He is not naive, and he is not alone. The argument has serious backers.

Satya Nadella has said that AI agents such as Copilot will take over the work of using programs like Excel and Word, with users simply giving prompts. Elon Musk went further on the Joe Rogan podcast in October 2025, predicting that within five or six years phones will have no apps and will simply display whatever an AI anticipates you want. Neither has shipped that future yet, and both are describing a direction rather than a product. But the direction is consistent. You describe the tool you need, and the tool appears.

Vibe coding, the practice of describing software in plain language and letting AI write it, is how that idea reaches a bank first. It is also the version most likely to be tested, because it needs no one's permission to start. Follow the head of innovation's logic and it leads somewhere uncomfortable for fintechs. If a feature can be produced in an afternoon, a feature stops being worth paying for.

I think we are further from that point than the prediction suggests. At a recent roundtable, we asked 20 senior people to rate, from one to ten, how far AI is embedded and delivering results in their organisation. The average was six. That number hides the interesting part. Vendors were in the room and scored themselves a ten. Some financial institutions gave themselves a one. The distance between those answers is the real finding. Vendors describe what their technology can do. Banks describe what their organisation can absorb. The second question sets the pace.

That does not mean the shift is imaginary. I know a payments fintech that raised its development speed twentyfold within a couple of months and automated its entire customer service operation with AI models and vibe coding. The technology works. But a payments fintech with a focused product and a small team can change direction in a week. A regulated institution of real size carries security reviews, model risk committees, data access rules, audit trails, DORA obligations and the EU AI Act. Writing the code is the fast part. Getting it approved, owned and maintained is the slow part.

So if building becomes cheap, what is a fintech for? The old answers were agility and a better user experience. Both are capabilities, and a bank with cheap software can acquire both, so neither survives as a definition. What remains is harder to prompt into existence. A fintech serving dozens of institutions sees fraud patterns and failure modes that no single bank can see alone. A fintech holds licences, signs service levels and becomes the party a regulator or a board can point to. A fintech goes deeper on one narrow problem while a bank manages hundreds of priorities.

I would name this change a move from selling software to sharing responsibility. The value is no longer in handing over a finished tool. It is in carrying part of the knowledge and part of the risk alongside the bank. We have already seen this pattern in RegTech, where banks and specialists work together because compliance is too large and too fast-moving for one institution to carry. That collaboration is not a future scenario. It is how the industry already behaves where the stakes are highest.

For fintechs, this changes the sales conversation. A feature demo will lose to a banker with a prompt. A conversation about who holds the risk, whose data improves the model, and who answers to the regulator when something breaks will not. Fintechs that can show what they know and what they are willing to be accountable for will be treated as partners. Those that cannot will be compared to a weekend project, and sometimes they will lose that comparison.

Vibe coding will not remove the need for fintechs. It will remove the need for fintechs that only sell features. For banks, the question changes as well. It stops being whether they can build it themselves and becomes whether they should carry it alone.