From Qonto to Well: What 7 Years in Fintech Taught Me About the Real Problem

It was 2 AM on a Tuesday in 2017, and I was staring at a screen full of SEPA XML messages that refused to validate. I was twenty-five, six months into building iBanFirst's core banking system, and I had just discovered that a single misplaced tag in a payment file could freeze €2M in cross-border transactions.

I fixed the bug, pushed the code, watched the payments clear, and felt the particular satisfaction of making money move correctly through pipes that didn't exist six months earlier.

That feeling — the builder's high of making infrastructure work — carried me through two and a half years at iBanFirst, where we shipped Europe's first fully digital onboarding for a payment institution. Twenty times cheaper than the paper process it replaced.

The lesson I thought I was learning: build better tools, and businesses will thrive. The lesson I was actually learning, which would take me five more years to understand: tools are never the bottleneck.

The Qonto Years

I joined Qonto in 2019 to build their core banking infrastructure. Nine months to ship a system that would eventually handle transactions for over 500,000 businesses. The pace was brutal and exhilarating.

Qonto was, by any reasonable measure, a success story. We scaled from a neobank with ambition to the leading business finance platform in Europe. I moved from building infrastructure to leading integrations — the connectors that linked Qonto to the rest of a business's tool stack.

Here's where things got interesting. When you're the integration lead at a company with half a million customers, you have a front-row seat to how businesses actually use their tools. And what I saw, consistently, across every segment and every market, was this: they don't.

Not really.

A business would sign up for Qonto. They'd connect their accounting software. They'd link their invoicing tool. And then, within the first week, 60 to 70 percent of them would stop using the integrations entirely.

Not because the integrations were broken — they worked fine. Not because the businesses didn't need them — they desperately did. They stopped because the cognitive overhead of maintaining yet another tool was higher than the perceived value of what they got back.

We'd ship a beautiful new integration — Slack notifications for payments, automatic receipt matching, CRM syncing — and the adoption curves all looked the same. A spike of signups, a cliff of churn, and a long tail of power users who would have figured it out regardless. I started calling it the Adoption Crisis.

The Board Room Epiphany

There's a moment that crystallized everything. I was in a board meeting — not presenting, just there to answer technical questions about integration metrics. A board member asked: "What category does Qonto compete in?"

The room went quiet. Not because nobody had an answer, but because everyone had a different one. Neobank. Business finance platform. SMB operating system. Each label implied a different strategy, different competitors, different success metrics.

I sat there thinking: the problem isn't that Qonto is in the wrong category. The problem is that categories built around tools are the wrong abstraction. A business owner doesn't wake up thinking "I need a neobank" or "I need an invoicing tool." They wake up thinking: How's my business doing? Where's my cash? Can I afford to hire next month?

Every tool on the market answers one slice of that question. No tool answers the whole thing. And the integration layer — the thing I'd spent years building — was supposed to bridge the gaps. But integration is glue. Glue holds things together; it doesn't think.

What businesses needed wasn't better glue. They needed intelligence.

The Cyclist's Lesson

I need to make a detour here, because the thing that finally pushed me to leave Qonto wasn't a business insight. It was a bike race.

I'm a competitive ultra-distance cyclist. I've won Race Across Paris twice and finished third in Race Across France — 350 kilometers through the night with no support car and no team. You and the road and the voices in your head that tell you to quit.

Ultra-distance cycling teaches you something that's hard to learn in an office: the difference between a problem of capability and a problem of endurance. In a 350km race, you have the legs to finish from kilometer one. What kills people is the accumulation of small frictions — a headwind that's just strong enough to be demoralizing, a climb that's just steep enough to make you shift one gear too many.

Business tools are like that headwind. Each one is manageable. Each integration is fine. But the cumulative friction of managing six, eight, twelve tools — that cumulative friction is what kills small businesses. Not any single tool. The stack itself.

I finished my second Race Across Paris thinking about this. Somewhere around kilometer 180, covered in rain, I had one of those clarity moments that only happen when your conscious mind finally shuts up. You don't solve a friction problem by adding more capability. You solve it by removing the friction.

The Intelligence Layer

I left Qonto in 2023. Not because I was unhappy — I loved the team and I was proud of what we'd built. I left because I couldn't unsee the problem, and I couldn't solve it from inside a bank.

The problem, stated plainly: every business tool is a silo. Integrations connect silos but don't eliminate them. The business owner remains the integration layer — the human who must hold the full picture in their head, reconcile conflicting data, and make decisions with incomplete information.

The solution: don't build another tool. Build an intelligence layer that sits on top of all the tools. Something that connects to a business's existing stack — whatever that stack happens to be — and turns fragmented data into a coherent picture. Something you can talk to.

Not a dashboard. Not a report. A conversation.

What I Actually Learned

Seven years in fintech taught me three things.

First: the tool layer is commoditizing. The SaaS market lost $285 billion in value in a single day in February 2026. That wasn't a correction — it was the market recognizing that AI will compress the value of traditional business tools faster than anyone modeled.

Second: adoption, not features, is the real crisis. I have the data from 600,000+ businesses to prove it. The 60-70% first-week churn isn't unique to Qonto. It's universal. The problem isn't that tools are bad. It's that humans are bad at using tools in combination.

Third: the conversation is the interface. When a business owner wants to know their cash position, they don't want to open three apps and cross-reference numbers. They want to ask a question and get an answer.

This isn't a feature request. It's a paradigm shift. The era of the tool stack — where humans are the integration layer between disconnected applications — is ending.

Building the Opposite

So now I'm building the thing I wish existed when I was at Qonto. It's called Well.

It's not a bank. It's not an invoicing tool. It's not a CRM. It connects to all of those things — and fifty more — and turns them into a business you can talk to. Every design decision is the opposite of what I would have done five years ago.

At Qonto, we asked: "What features should we build?" At Well, we ask: "What questions should a business be able to answer?"

The irony isn't lost on me. I spent years building the most sophisticated integration layer in European fintech, only to realize that the integration layer itself is the problem. The answer was never to build better connectors. The answer was to build something that makes the connectors invisible.

Seven years ago, I thought better infrastructure was the answer. I was wrong, and it took building infrastructure for half a million businesses to see it. The answer isn't better tools. It's making tools disappear into intelligence.

That's what we're building.

— Maxime Champoux, co-founder and CEO of Well