Headway is an Every Consulting client; we delivered its leadership team a paid executive AI workshop. Headway was given an opportunity to fact-check details but had no editorial control over this article.

Was this newsletter forwarded to you? Sign up to get it in your inbox.

At mental healthcare company Headway, an AI agent can act without asking permission at every step. That might sound unexpected for a 900-person company in an industry governed by strict rules for sensitive patient information. But the autonomy is possible because of Headway’s tight controls around the agent. Every conversation runs inside a sealed, disposable container, with carefully limited connections to company tools and data.

That architecture underpins Eddy, the internal AI assistant Headway built when existing products couldn’t meet its particular combination of security, compliance, and workflow requirements. Eddy is built on the Claude Code SDK, Anthropic’s toolkit for teams developing products with Claude Code. It’s hosted in Headway’s Amazon Web Services environment and connected to the company’s tools and data.

Work on Eddy had barely begun six months ago. Headway initially wanted an existing product and was preparing to sign a deal with a major AI vendor to give employees access to its desktop app. But relying on vendors to deliver features that met Headway’s workflow and strict compliance requirements for handling sensitive patient data “felt like we were missing the train,” says chief technology officer Arnaud Ferreri.