Interviews with Leaders in Fintech & Web3

The episode contains a handful of genuinely useful operational data points about Funding Circle's decision engine, but significant runtime is consumed by career biography, generic AI cheerleading, and career-advice segments for early-career listeners that add nothing for a B2B operator.

“our kind of risk discrimination is about three times better than if you were just to use a bureau to make that decision because we've got that additional uh, the benefit of that additional, that moat if you like of data that we have that others simply don't”

“We now have uh, a customer interaction every 38 seconds now. Right. Uh, and all of that feeds into our intelligence and enables us to open up new pockets of lending”

The 'product engineers not software engineers' framing and the AI-Native vs AI-First distinction are genuinely interesting conceptual moves, but the bulk of the episode recycles standard AI optimism and generic career advice without any contrarian or first-principles argument.

“we call our team like product engineers. We don't have software engineers. Right. And that's an important distinction that we make. And what that means to me is that the role is to um, create products to solve problems, right? It's not to develop software.”

“AI first to me is, um, that should default. You go to AI for every solution. Uh, and I don't think that's necessarily the case.”

Greg McEwen is a credible practitioner - listed-company CTO with prior scale experience at Paddy Power Betfair across 100 countries - and speaks from genuine operational experience building a ninth-generation proprietary credit model; however his answers stay at a comfortable altitude and rarely reveal genuinely proprietary thinking.

“We are on our 9th generation of AI risk uh model now as well. So that is the model that we have built in house over the many years that we've been operating”

“we have 15 years worth of uh, lending data ah, uh, from the products and services that we've offered that we can use to augment that decision making process”

Several concrete numbers land well - 9th-generation model, 15 years of data, 3× discrimination vs bureau, 38-second interaction cadence, 70%+ instant decisions - but the episode's most tantalising claim (a three-day Claude Code prototype that would normally take years) is deliberately left undescribed, and many strategic statements are vague assertions without supporting data.

“70% or more actually of applicants will get an instant decision within seconds”

“our kind of risk discrimination is about three times better than if you were just to use a bureau”

The host lands one genuinely sharp definitional follow-up ('How would you define AI Native versus AI First?') and probes the evolution of the risk model, but allows sweeping claims like '3× better discrimination' and 'phenomenal' Claude Code output to pass completely unchallenged, and spends a disproportionate share of the interview on career biography and early-career motivation content.

“And how's that changed? I suppose in like you said the ninth generation before Genai was using machine learning and more traditional approaches. And is it constantly evolving?”

“How would you define AI Native versus AI First?”