Relay is shutting down, and the founder is walking into Google.
TechCrunch reports that Relay, the AI workflow automation tool that launched in 2021 with the ambition of becoming the new Zapier, is ending access for free customers as of August 15 and for paying customers on September 14. The closure was first announced in July. Jacob Bank, Relay’s founder and CEO, is rejoining Google as VP of Product for Chrome, where he will lead the product and developer relations teams.
Bank spent a little over six years at Google before, working on Gmail, Google Calendar, and Google Chat, after his scheduling startup Timeful was acquired by the company in 2015. He is going back to the place that just absorbed his second company’s team.
That is the part worth sitting with.
The agent layer is being absorbed
Relay sold a very reasonable idea: a tool that drafts documents, edits copy, and automates repetitive project work so a person can get more done without giving up control. The problem is that the same capability is now a feature, not a company.
Chrome is the surface where a huge share of that work already happens. Bank put it plainly in a post on X, calling Chrome “a perfect place to collaborate with agents” and saying he has “really ambitious plans to help you work with AI in Chrome to get things done.” Gemini is already an optional in-browser assistant, and Google recently reported that Gemini has crossed one billion users.
When the platform that hosts your work decides to do the work for you, the standalone automation layer has a hard time explaining why it exists.
I have written about coding agents and constraint decay before, about how agent products keep losing ground to the environments they run inside. This is the same pattern at the product level. The interesting capability does not stay in the tool. It migrates to the place where the user already is, and the tool becomes a feature with a changelog.
None of this is a verdict on Relay. The company shipped real things, and the team is not disappearing, it is being absorbed, which is a better ending than most startups get.
But the signal is clear. The next round of agent products will be judged less by what they can do and more by where they sit relative to the surfaces people already live in. If that surface is the browser, the platform is not going to wait for permission to build the agent into it.
Source: TechCrunch