Startupeable published an episode on July 15 with Pedro Pineda, co-founder and CEO of Fintual, the Chilean investment-management fintech. The number that opens the conversation is blunt: Fintual manages more than $2.2 billion in funds, double what Nubank’s funds manage. With a customer base of just 200,000, against Nubank’s 100 million.

Pineda’s explanation is simple: managing money isn’t a bank’s core business. Lending it is. Fintual was built to be the entity whose only job is the part banks treat as a side business.

That contrast alone would carry a post. What made me write this one is what Pineda says he did after realizing something that, by his own account, cost him sleep: a traditional bank would never beat Fintual on its own turf. But a new company, AI-native by design, could beat Fintual.

The decision: build your own replacement before someone else does

Most founders defend what they built. When something works, the instinct is to protect it, optimize it, shield it from competitors. Pineda did the opposite. He built an isolated team inside Fintual, separate from daily operations, with an explicit mandate: rebuild Fintual from zero, using AI, as if it were the competitor coming to eat the business.

That bet let him recruit a Google engineer as the new CTO, someone who joined specifically to lead the internal rebuild, not to maintain the existing system.

Three years later, per Pineda, the result shows up everywhere inside the company. Fintual’s lawyers write code. Not as a showcase innovation exercise, but because it’s now part of how they do their jobs. And an internal team just eliminated a critical customer-support software vendor, building the replacement themselves in one week.

That last detail is the one people get wrong.

Why building it in a week isn’t the same as building it fast

It’s easy to read “rebuilt the replacement in a week” and assume AI did the heavy lifting alone. That’s not what happened, by Pineda’s own telling. A team can only rebuild a critical customer-support system in seven days if it already knew, precisely, what that system had to do, which edge cases it had to handle, and exactly where the outside vendor fell short for their specific operation.

AI didn’t replace that knowledge. It executed it at a speed that didn’t exist before. The bottleneck was never writing the code. It was knowing exactly what code to write. That’s the same pattern I wrote about a few months back in a post on why building got cheap while deciding what to build didn’t: once execution cost drops to near zero, what separates one company from another is the quality of the decision about what deserves to exist, not the speed at which anyone can type.

Fintual could make that call fast because the team making it was the same team that had spent years running the process it was replacing. This wasn’t an outside consultancy reading a manual. It was people who already knew where the bodies were buried.

”AI already knows more about my company than I do”

That’s the line Pineda uses to describe where the experiment landed. It isn’t a marketing metaphor. It’s a specific admission: after three years of feeding internal systems with the full context of how Fintual operates, AI now has access to more surface area of the operation than any single person, including the CEO, can hold in their head at once.

That only happens because Pineda didn’t treat AI as a tool bolted onto an existing process. He treated it as the foundation he rebuilt the entire process on top of, with the original team’s operational knowledge as the raw material, not something AI had to guess from scratch.

The next bet: pension funds

Pineda isn’t satisfied with having doubled Nubank’s fund AUM with a fraction of its customers. His stated goal is to do to pension funds across Latin America what Nubank did to traditional banking: enter as a technology-native alternative to a sector dominated by large, regulated, slow-to-modernize institutions.

It’s a bigger bet than the last one, and there’s no result to show yet. But the logic is the same one that already worked once: use the internal AI rebuild as a structural advantage, not as a pilot project kept isolated from the core business.

What this means if your company isn’t a Chilean AI startup

The Fintual case stands out precisely because it isn’t a Silicon Valley story. It’s a Latin American company that decided to disrupt itself before someone else got the chance, and documented it with verifiable numbers instead of a vague promise of “digital transformation.”

The part almost nobody replicates well is the one Pineda mentions almost in passing: the team that rebuilt the process already understood it deeply. In AI Maestro, discovery exists to reach that exact starting point before a single line of code gets written: mapping the real operation, not the one described in the procedures manual, so the AI rebuild starts from genuine process knowledge instead of an optimistic guess about what the tool can figure out on its own. I wrote before about other Latin American companies that have this same opportunity in front of them and haven’t taken it yet.

Map your process before you rebuild it with AI