Coinbase chief executive Brian Armstrong says the exchange now runs parts of its business through AI agents that operate at the level of L5 or L6 employees. In a Nov. 18 interview with David Senra, he described a shift from writing prompts to assigning goals, with agents leading teams of humans or other agents and asking for input on decisions that matter.
The structure he described is seven or eight lead agents, each responsible for a different area of the business, coordinated by a chief-of-staff agent. Each lead can deploy three to five subagents while it works. Armstrong called the setup an unlimited team of L5-level staff, and compared running a company to strategy games where progress comes from coordinating people rather than doing every task yourself.
Humans still review important decisions and keep responsibility for direction and approvals, with a prioritised inbox used to surface which agent decisions and projects need attention. Coinbase is also building an internal system it calls “Brain” that collects company knowledge — code reviews, research, incident reports, meeting records — so that agent output matches how the firm already does things.
Armstrong was candid about the limits. Current systems can still miss context or produce derivative work, which is a meaningful caveat for a company whose product is custody and execution, where a confident but wrong answer is expensive. The goal, he said, is substantially higher productivity per employee rather than replacing the judgement that sits above the agents.
His larger thesis is what he calls an agentic economy, in which AI agents transact and collaborate across company boundaries. In that model crypto supplies the payment rails, stablecoins in particular, while AI supplies the interface. He argued the two are complementary rather than competing, and that agents could eventually help people invest against stated goals or handle routine financial tasks, giving ordinary users something closer to a personal banker.
He also put a number on the opportunity: fintech captures roughly 4% of global financial-services revenue, leaving most of the market on traditional rails. Armstrong’s view is that closing that gap needs products that are dramatically better rather than marginally cheaper, and that AI-driven services are the likeliest reason a customer would switch.