Founder Dossier No. 102 · 5 min read
Pedro Franceschi
Franceschi rebuilt Brex's internal security model so its AI agents could operate with fewer manual restrictions, applying an automated judge at the network boundary rather than the tool layer, then negotiated Brex's 5.15 billion dollar sale to Capital One in about 40 days.
As a teenager in Rio de Janeiro, Pedro Franceschi reverse-engineered Siri to speak Portuguese and got a legal notice from Apple for it. In January 2026 he agreed to sell Brex, the company he built after that, to Capital One for 5.15 billion dollars, in a process he says ran about 40 days from first conversation to signed agreement.1
Franceschi is the co-founder and CEO of Brex, a financial technology company that provides corporate credit cards, banking, and spend management software to startups and larger businesses. He built Brex with fellow Brazilian founder Henrique Dubugras after a teenage career in software and payments, and in 2026 led the company through its sale to Capital One.
Background
Franceschi was born in Rio de Janeiro in 1996. His father, a computer enthusiast who owned a rare home computer for Brazil at the time, developed lymphoma and died when Pedro was eight; his mother, a psychologist who later moved into corporate marketing, raised him and his younger sibling alone afterward. Franceschi connects his early absorption in computers to that loss, describing programming as a way to "insert yourself in a place where you control all the variables" when the outside world offers none.1
He taught himself C and C++ around age eight or nine, learning English at the same time, and was hired by a mobile app company around eleven or twelve. Under the handle PH, he became one of the first people to jailbreak the iPhone 3G and later reverse-engineered Siri into Portuguese, drawing a legal notice from Apple, then built and sold Quazar, a multi-window iPad app that brought him roughly the equivalent of hundreds of thousands of dollars as a teenager.1
Getting started
Franceschi's path into payments was accidental. Around age fourteen he took a job at a Brazilian startup modeled on Square to fix an iOS security bug, and saw firsthand how outdated the country's card infrastructure was. In late 2012 he met Henrique Dubugras, a Brazilian teenager chasing the same problem from the merchant side; the two became co-founders after meeting on Twitter and, in 2013, founded Pagar.me, a payments processor they later realized resembled Stripe, a company neither had heard of. Pagar.me built its own processing stack down to direct Visa and Mastercard connections under central bank regulation, with its teenage founders as licensed officers, growing past a hundred employees and a billion dollars in processed volume before Stone, a larger Brazilian rival, acquired it around 2016, funding both founders' move to the United States.1
Franceschi enrolled at Stanford at Dubugras's urging; neither finished more than a few quarters. In January 2017 the pair entered Y Combinator's winter batch with a virtual reality idea, then pivoted within weeks to a problem they knew from Pagar.me: venture-backed startups with millions in the bank were routinely denied corporate cards for lack of revenue history. Brex was built to issue cards against cash position instead.1 The name was simply an available four-letter domain; the "Brazilian Express" story came later.
What he built
Brex launched its startup corporate card in 2017, reached unicorn status within about two years, and later expanded into mid-market and enterprise finance. It built its own card-issuing infrastructure directly on Mastercard's rails rather than licensing a third-party platform, later extending that stack to over a hundred countries. By the mid-2020s Brex served more than 35,000 customers, including over 300 public companies, and had raised roughly 1.5 billion dollars across eleven rounds, reaching a 12.3 billion dollar valuation in January 2022.1
Brex went through a difficult reset in 2023 and 2024, a period Franceschi calls "Brex 3.0," cutting about 30 percent of staff, removing two management layers, and repricing employee equity down from the 2022 peak. In January 2026, Capital One agreed to acquire Brex for 5.15 billion dollars, closing mid-2026 with Franceschi remaining as CEO; he has said it moved from first conversation to signed agreement in about 40 days, crediting Capital One CEO Richard Fairbank's approach of preserving Brex's independence.1 Franceschi frames the deal as a bet that traditional corporate cards still cover the large majority of a large and growing U.S. spend market, and that Capital One's far larger marketing and R&D budgets give Brex a faster route to that share.
How he operates
Franceschi has built a reputation as an unusually hands-on, technically engaged CEO who keeps writing code and shaping product decisions well past the point where most executives step back from both. He believes a CEO cannot lead product development credibly without understanding the frontier of the technology itself, and that only the CEO can resolve cross-functional conflicts in minutes that would otherwise take a team days.
At Brex, Franceschi pushed the company to give AI agents real operating autonomy rather than keep them on a short leash. To make that workable inside a regulated financial company, Brex built an internal system called CrabTrap that records an agent's normal traffic for a day, then routes anything outside that pattern to an automated reviewer at the network level rather than restricting the agent's tools directly; Brex says most internal agent requests now clear automatically.2 Franceschi runs much of his own workload, including document review, message drafting, and follow-up tracking, through a personal AI system he built and named Lemon Pie, and says roughly a third of his personal purchases are now agent-made.3 He believes a genuinely new product is best introduced through a small, separate team rebuilding the experience from scratch, since existing processes are built to resist change, and has spoken publicly about burnout in his twenties, crediting therapy, meditation, and a household rule against AI on Saturdays.
Where things stand
As of mid-2026, Franceschi remains CEO of Brex, now part of Capital One, with plans for the company to keep meaningful independence inside its new parent. He has said he wants Brex to outlast his own tenure the way Pagar.me remains one of Brazil's largest online payment processors a decade after he left it. He continues to be publicly associated with heavy personal AI use, both as a builder of internal Brex systems and a daily user of his own agent tools.
Key facts
- Franceschi taught himself to program around age eight or nine and became one of the first people to jailbreak the iPhone 3G as a teenager in Brazil.
- He co-founded Pagar.me with Henrique Dubugras in 2013 as a teenager; Stone, a larger Brazilian rival, acquired it around 2016.
- He and Dubugras founded Brex in Y Combinator's Winter 2017 batch, pivoting within weeks from a virtual reality concept to a corporate card for startups without credit history.
- Brex reached a 12.3 billion dollar valuation in January 2022 before a 2023-2024 restructuring that cut roughly 30 percent of staff.
- Capital One agreed to acquire Brex for 5.15 billion dollars in a deal signed January 2026 and completed mid-2026, with Franceschi remaining as CEO.
- Franceschi built and uses a personal AI agent, Lemon Pie, for much of his workload, including roughly a third of his personal purchases.
This subject remains under active examination by the institution. The file enters the general collection when the dossier is complete.
References
- 01
He's Built The First Full-Time AI CEO (Pedro Franceschi, Core Memory Podcast)
Pedro Franceschi, interviewed by Ashlee Vance · podcast · 2026-03
- 02
The Most AI-Pilled CEO We Know
Pedro Franceschi · podcast
- 03
The Agentic Commerce Revolution
Pedro Franceschi · interview
From the Curator
The catalog continues with the file on Phil Knight, Dossier No. 105.
Founder Dossier No. 105Phil KnightRefused ever to cut an inventory order, doubling it after every sellout and signing over his own house as collateral, running perpetually near bankruptcy because to him pulling back was simply losing.Also on the desk: Enterprise Coopetition (Concept practiced)
Related