Brian Armstrong, CEO of Coinbase, announced that the exchange has launched a test run of AI agents working inside the company alongside real employees — showing up in Slack and corporate email, receiving tasks, and delivering results.
The first two agents are modeled after former Coinbase executives. Fred Ehrsam, in AI form, became a "strategic executive agent" — handling priorities, documents, and feedback. Balaji Srinivasan was turned into an "agent of chaos and creativity": challenging assumptions, developing long-term concepts, generating unconventional ideas — everything the real Srinivasan is known for in the industry.
Armstrong also outlined where the project is headed next.
"Employee agents shouldn't be digital twins of specific people — each one should have its own name, its own identity. That's the next step," he said.
In the comments, some users praised the initiative. Others pointed out that a name is just the surface, and deeper problems lurk underneath.
"Who is responsible for an agent if it makes a mistake or leaks customer data? These things hold up fine until the first routing error," one user noted.
Skeptics had plenty to say too.
"Will named agents help find the $667M the exchange lost last quarter, or are they just decoration while management dumps another 40,000 shares?" another user wrote sarcastically.
Armstrong, meanwhile, sketched out a broader vision: every Coinbase employee should be able to spin up agents for any task. While the exchange tests its first AI employees, the industry conversation has already moved further — from names to wallets, accountability, and whether an agent has the right to its own business model.