GitHub announced shared Copilot Automations on 28 October 2026. An automation combines a prompt, a trigger and a set of allowed tools, and Copilot runs it in the cloud. GitHub’s examples are reproducing a bug when an issue opens, and investigating a production incident when a monitoring tool raises one.
Until now, automations were personal. They were visible only to their creator and used that person’s identity and AI credits. Shared automations change four things.
- Ownership: the automation belongs to the repository. Anyone with read access can see it, anyone with write access can edit it, and it keeps running if its creator leaves.
- Billing: AI credits are charged to the organisation or user that owns the repository, not the person who built it.
- Attribution: actions are attributed to Copilot, not an individual.
- History: a change log shows how the tools, triggers and prompt have changed over time.
For organisation-owned repositories, an administrator must first enable the policy allowing Copilot cloud agent billed to the organisation. The type is chosen at creation (Shared) and cannot be changed afterwards. It is rolling out to Copilot Pro, Pro+, Max, Business and Enterprise plans.
What it means
This is the point where AI automation stops being a personal productivity trick and becomes part of how a codebase is run. That is useful, because the workflow survives staff changes. It also means an AI agent may be acting on your repositories, on a trigger, with spend attached, without a person pressing a button.
So what
If an agency or in-house team works in your repositories, ask which automations exist, what tools they are allowed, who can edit them and who pays. Those answers belong in the working agreement. Our AI-assisted development service sets these controls up from day one. To discuss your set-up, start a project.