Audit-ready AI agents doing the reading, sorting and citing so bank compliance teams only touch the cases that matter.
A logo on a plate, a name borrowed from 1944. The startup once called Greenlite decided its work - keeping money clean - deserved a bigger namesake. So it took one from the agreement that built the modern financial system.
Banks employ a lot of people to read alerts. A transaction trips a rule, someone in a back office pulls up the customer, checks the history, decides whether it is money laundering or a false alarm, and writes it down. North American institutions spend north of $60 billion a year on this - and most of it is exactly the sort of high-volume, procedure-heavy reading that software is supposed to be good at. Bretton AI's whole argument is that this work is the most automatable job in banking, and somehow still the least automated.
So Bretton builds AI agents that plug into a bank's existing compliance stack - the detection systems and case-management tools already in place - rather than asking anyone to rip anything out. An agent ingests the alert and the bank's own policies, pulls together the evidence scattered across systems, closes the low-risk cases on its own, and escalates the genuinely complicated ones to a human with a narrative that is already written, cited and audit-ready.
The company was founded in 2023 as Greenlite AI, went through Y Combinator that summer, and spent two years and three funding rounds building toward a rebrand. In February 2026 it became Bretton AI - after Bretton Woods, the 1944 conference that set the rules for the post-war financial system. It is a large name to pick. The company's answer is that the AI era needs a new trust layer for finance, and it would like to be it.
Whether or not you buy the framing, the numbers underneath are real: agents that have made more than 1.2 million investigation decisions, roughly 195,000 hours of manual work eliminated, and a roster of regulated banks and fintechs that includes names like Mercury, Gusto and Upgrade.
Rather than ripping and replacing existing detection or case-management systems, Bretton AI plugs directly into the current compliance stack. — Sapphire Ventures, on why it led the round
Regulated banks run compliance, risk and fraud on one system using their own data and policies. Agents complete L1 and L2 investigations in minutes instead of days - across AML, KYC, KYB, sanctions and transaction monitoring.
The part regulators care about: model risk management, continuous AI evaluations and quality-assurance testing. Every output is traceable and cited, so an agent's decision can survive an exam.
AI coworkers that ingest alerts, consolidate evidence across systems, auto-resolve the low-risk cases and escalate the hard ones with audit-ready narratives already written for the human reviewer.
OCC-, FDIC- and Federal Reserve-regulated banks alongside fintechs. Named customers and partners include Mercury, Gusto, Upgrade and Paxos. Across the book, Bretton says it serves institutions representing more than $1 trillion in total market capitalization.
Grow the bank without growing the back office. — the CFO-facing version of the pitch
Vertical AI lives or dies on whether the agents have actually done the work. Bretton's headline figures, drawn from its own disclosures and its lead investor's write-up:
Reported reductions from named case studies. Bars scaled to each metric's own range.
Sources: Bretton AI Series B announcement and Sapphire Ventures. Figures are company-reported and approximate.
Led product for anti-money-laundering and payments compliance at Meta (Facebook / WhatsApp), then built core compliance infrastructure at Paxos, the crypto-infrastructure firm powering millions of wallets. His investors' one-line summary: the right founder for this category, because he lived the problem first.
Co-founded the company in 2023 and leads engineering. Together with Lawrence he built the team's core thesis: that governance and auditability aren't features bolted onto a compliance product - they are the product.
Will Lawrence and Alex Jin start the company and join Y Combinator's Summer 2023 batch.
Capital to bring AI coworkers to bank and fintech compliance teams.
Greylock and Thomson Reuters Ventures back a trusted AI compliance workforce.
Sapphire Ventures leads; the company takes its name from Bretton Woods and expands across financial-crime domains.
"Bretton is an homage to Bretton Woods - the agreement that helped define the modern financial system."
"It was immediately clear he is the right founder to build the category-defining platform for financial crime operations."
"Financial institutions in North America spend over $60B on financial crime compliance costs annually."
| Legal name | Bretton AI, Inc. (formerly Greenlite AI, Inc.) |
|---|---|
| Founded | 2023 · Y Combinator Summer 2023 |
| Headquarters | San Francisco, California, United States |
| Founders | Will Lawrence (CEO), Alex Jin (CTO) |
| Category | AI · Fintech · RegTech · B2B SaaS |
| Focus | AI agents for AML, KYC, KYB, sanctions and transaction monitoring |
| Team size | ~70 employees |
| Total funding | $90M+ (Seed, Series A, Series B) |
| Latest round | $75M Series B · led by Sapphire Ventures · Feb 2026 |
| Named customers | Mercury, Gusto, Upgrade, Paxos |
It provides an AI-native platform whose agents run compliance, risk and fraud operations - AML, KYC, KYB, sanctions and transaction monitoring - for regulated banks and fintechs, with every output cited and audit-ready.
Yes. It was founded in 2023 as Greenlite AI and rebranded to Bretton AI in February 2026 alongside its $75M Series B. The name references the 1944 Bretton Woods agreement.
Will Lawrence (CEO), who previously led AML/compliance product at Meta and built compliance infrastructure at Paxos, and Alex Jin (CTO).
Over $90M total - a $4.8M seed, a $15M Series A (2025) and a $75M Series B (February 2026) led by Sapphire Ventures, with Greylock, Thomson Reuters Ventures, Canvas Ventures, Y Combinator and TIAA Ventures participating.
Through its Trust Infrastructure - a governance layer with model risk management, continuous AI evaluations and quality-assurance testing that makes every agent output traceable and cited for regulators.