Rewiring Fintech: Lessons From Paddle, Checkout, and Dust

On July 7, we brought together a room of 80 fintech operators in London for a panel that was supposed to last 40 minutes. We ran over (and it’s usually a good sign).

Moderated by Seb Johnson (Scaling Europe), the conversation brought together Hendrik Brackmann, VP of Data & AI at Paddle; Andy Cornforth, Senior Engineer at Checkout.com and Faateh Dhillon, UK Lead at Dust.

The question on the table: what does it mean to use AI to rewire how a fintech operates, not as a product feature, but as the way the company runs?

Here's what they said.

Before anything else, the panel had to set the ground into defining what an AI operating system means. We hear this concept often, with less precision than it deserves.

Dust framing was simple: the goal is to get context flowing across the organisation. "The real value isn't the AI model," as Faateh said. "It's making sure the right information reaches the right person at the right moment, across functions, without anyone having to ask."

An operating system, in that sense, isn't a piece of software. It's the condition under which everyone in a company can build on shared knowledge. That framing landed differently depending on where you sit. For Paddle, this means focusing AI on the core of each team's function - not the peripheral tasks, the ones that actually drive outcomes.

For Checkout, it started with infrastructure. "Before you can talk about intelligence, you have to talk about the data layer underneath," as Andy said. "We had to change what the infrastructure looked like before any of this became possible."

Every panelist was asked: what does getting a company to actually use AI look like? None of them said "a launch announcement."

For Paddle, "it's a journey. It starts with experimentation." The company’s approach wasn't to deploy a company-wide AI strategy from day one. The goal was to let teams explore and test on their own, in ways that felt safe and relevant to their processes. "The reputational model works better than a top-down mandate," he said. Peer visibility does the pulling: teams get inspired by the work built by others. This bottom-up approach appeared to create more sustainable results than a top-down one.

Checkout's view was starker. "If there's no ROI, it's not going to land." Andy was clear that at a company operating at Checkout's scale and under its compliance pressures, enthusiasm alone doesn't move the needle. Adoption has to show up in outcomes: faster decisions, fewer errors, time saved on things that used to eat hours.

Dust’s take was about mindset: "The mental shift is that everyone can build. Not just engineers, not just data teams but everyone." Dust's role is to empower individuals with the tools to achieve this shift real. Making it happen depends on a variable companies often underestimate: "Company culture matters. It is not a technology problem, but a cultural one."

The room expected a long section on compliance, what they got instead was a reframe. Hendrik's point is that "Financial services, as an industry, has one of the biggest advantages when it comes to moving fast.”. Indeed, Fintech companies are used to operating in regulated environments, they understand how to document decisions, trace outcomes, and build accountability into their processes. That's not a constraint on AI adoption, it's a head start.

Faateh furthered this point from an auditability angle: "You have to make sure your agents are delivering consistently and traceably across the company. That's not unique to fintech, it's just good practice." Auditability isn't the enemy of velocity: it's the catalyst making it sustainable.

The harder conversation was around open-source models and data security. "How do we keep our data safe?" is the question that comes up in every enterprise AI conversation, and fintechs feel it more acutely than most. Andy's framing was honest: "There's a tradeoff between speed and safety, and we haven't fully resolved it. The work right now is about being able to govern it better and to put the right practices in place before something breaks."

Engagement peaked when Seb pressed all three panelists on what wasn't working, and the room responded by taking notes:

Hendrik: "We've had agents fail because people didn't change their practices alongside them. Technology doesn't transform itself, people have to change too." One specific learning stick: agents need access to the right data to be useful. "We learned that access was the constraint, not the model."

Andy: "Speed and safety are genuinely in tension. We've moved fast and had to walk things back. The governance piece isn't bureaucracy, it's what makes you able to move fast sustainably."

Faateh: "Garbage in, garbage out. The single most common reason an AI implementation underperforms is bad input data. And underneath that is a harder question: is the data even available to the people who need it? A lot of companies think they have a model problem when they have a data access problem."

The Q&A ran long, with a room full of people asking how to go further: how do you get the third or fourth team to adopt what the first team built? How do you govern at scale without slowing down? How do you know if the culture shift has taken place?

The honest answer from all three panelists converged on the same point: the technology stopped being the constraint a while ago. What separates fintechs pulling ahead isn't which model they use or how sophisticated their stack is. It's whether they've built the conditions for knowledge to move, across teams, across functions, without friction.

That's what rewiring actually means. Not a bigger AI budget. A company that operates differently because information reaches the people who need it, when they need it.

More events like this are coming. If you want to be in the room, watch this space: Luma Dust London