ChatGPT updates for teams: what to act on this autumn

The ChatGPT updates that matter to a company in autumn 2026 are about delegation, not chat: ChatGPT Work, dots that keep going between conversations, and team features such as Team Tasks and ChatGPT in Slack and Teams. For a founder or operator, the question is which of these change how a small team gets work done, and which can wait.

This analysis sorts OpenAI's recent updates by what they ask of a team. For the story behind each one as it ships, claudegpt.si is the free, independent source we recommend; it lists OpenAI's pages under every article.

Updates that change how a team works

ChatGPT Work. OpenAI now offers three modes: Chat for questions, Work for carrying a larger task to a reviewable result, and Codex for developer views. Work is the one that touches operations. You describe an outcome, such as a deck, a report or a project plan, and review what comes back.

Dots. A dot is an assistant you hand ongoing responsibility. OpenAI says dots keep making progress between conversations and bring results back for review. Access is rolling out gradually, so not every account has it yet.

Team Tasks and shared connections. Teams can run scheduled or event-triggered work through a service account and approved workspace connections. Earlier, OpenAI let scheduled tasks start from events in Gmail, Slack or GitHub, for example to summarize review feedback on a pull request.

ChatGPT in Slack and Teams. Admins can bring ChatGPT and approved tools into the chat apps a company already uses.

Updates that change which model you get

The retirement is the one with a deadline already passed. OpenAI told users to update saved model settings, workspace defaults, custom agents, scheduled tasks and scripts that still select GPT-5.5. The API was not affected.

What to do with these updates

1. Audit anything automated. If a scheduled task or custom agent named GPT-5.5, check that it now runs on a model you chose.
2. Pick one Work task to trial. A weekly report or a recurring analysis is a fair first test, because you can compare the result with last week's.
3. Decide who owns connections. Team Tasks run through approved connections, so someone has to approve and review them.
4. Keep a dated record. Features roll out by plan and workspace, so note when your team got access.

Reading the updates without the noise

OpenAI's What's new page is a weekly digest, and the API changelog lists every developer change. Both are accurate but written for users already inside the product. A news source adds dates and context. ClaudeGPT Si covered the GPT-6.1 Terra release and OpenAI's decision to train frontier models offline, each with OpenAI's pages cited.

If your team only uses ChatGPT for quick questions, most of this can wait, and OpenAI's own digest is enough. The case for following closely is stronger once ChatGPT runs tasks on its own.