On September 25, 2026, GitHub announced an update to Copilot for Slack and Microsoft Teams. The assistant now draws on richer conversational context, lets users switch AI models mid-conversation, and handles long-running tasks more reliably. The feature enters public preview for Copilot Business and Copilot Enterprise organizations, rolling out gradually across workspaces.
Richer context for sharper answers
In Slack, Copilot can now pull in files, attachments, and message links from a thread to build its response. In Microsoft Teams, it reads embedded images, forwarded messages, and thread history. In both cases, Copilot automatically checks for similar tickets or issues before suggesting a new one, and provides a direct link back to whatever work item it creates, with traceability to the source conversation.
This mirrors a pattern already visible with the agentic coding push inside GitHub Copilot: the more application context an assistant can reach, the more autonomous it becomes, which raises the stakes on permission scope and attack surface.
Switching models without leaving the thread
A notable addition lets users switch AI model for their next message directly from a Slack or Teams thread. In Slack, teams can also set default repositories and owners per channel, which simplifies multi-repo work in conversations shared across teams.
Switching models on the fly, without leaving the conversation, is the real usability leap in this update: the trade-off between speed and answer quality becomes a decision made mid-conversation, not upfront.
Availability and rollout scope
| Criterion | Slack | Microsoft Teams |
|---|---|---|
| Status | Public preview | Public preview |
| Required license | Copilot Business or Enterprise | Copilot Business or Enterprise |
| Context used | Files, attachments, message links | Embedded images, forwarded messages, thread history |
| Default settings | Per-channel repositories and owners | Not specified yet |
| Rollout | Gradual, by workspace | Gradual, by workspace |
Reliability on long-running tasks
GitHub says it improved handling of long-running tasks and multi-step implementation plans triggered from Slack or Teams. In Teams, history retention and converted-image handling are described as more reliable. Dropped connections and abandoned sessions are also better handled, a sensitive point whenever an agent works across several consecutive messages without constant supervision.
A simplified example of what a default-repository policy looks like, as documented for Slack:
{
"channel": "#backend-team",
"copilot": {
"default_repository": "org/backend-api",
"default_owner": "org",
"model_override_per_message": true
}
}Key takeaways
This update does not overhaul how Copilot is used, but it removes friction for teams already running their work through Slack or Teams: fewer round trips back to GitHub, more context captured automatically, and a model choice that adjusts on the fly. The gradual rollout, however, means the feature will take several weeks to reach every eligible organization. The same logic of reducing daily friction is visible on the tooling side, where teams are also re-evaluating their JavaScript package manager choices for the same reasons: less waiting, less manual coordination.
What this kind of integration really changes is how non-technical teams perceive development work: a ticket created automatically from a Slack message, with a traceable link, makes an assistant’s work visible without anyone needing to open GitHub. That is worth watching for teams still tracking their projects by hand — Simon Janvier.
Primary source: GitHub Changelog, September 25, 2026.
