Your Markdown Runbooks Can Now Work in Copilot Notebooks

A lot of useful IT knowledge does not live in a beautiful Word document. It lives in README files, Markdown runbooks, exported logs and the text file somebody updates because it is faster. Copilot Notebooks can now work with those files directly.

Microsoft announced support for Markdown, plain-text and rich-text references on 13 August. The feature began rolling out to Copilot Notebooks users that week.

A useful example: runbook versus incident log

Imagine you have two sanitized files:

• mail-flow-runbook.md contains the approved troubleshooting sequence for a fictional mail relay;
• incident-2026-08-14.txt contains a test incident timeline with timestamps and actions.

Add both as references to a Copilot Notebook and ask:

Compare the incident timeline with the troubleshooting sequence in the runbook. List the steps that were completed, skipped or completed out of order. Cite the relevant part of each source. Do not infer a step unless the incident log explicitly records it.

That is much more useful than asking for a generic incident summary. The notebook has a defined source set and the question tells it what evidence matters.

Check the citations before using the conclusion

If the answer says a step was skipped, open the cited source. Was the step genuinely absent, or did the log use different wording? If the runbook contains alternatives, did the answer mistake them for mandatory steps?

This is especially important with operational documentation because a confident but wrong conclusion can send the next person down the wrong troubleshooting path.

Use the result as an analysis aid, not a replacement for the engineer responsible for the incident.

Keep source files current

Adding a Markdown file to a notebook does not magically turn your Git repository into a live synchronized knowledge base. Treat the reference as the version you added and verify the update behaviour for your specific workflow.

If the runbook changes frequently, define who refreshes the notebook reference or use an integration designed for the source system. Otherwise users may be asking good questions against last month's instructions.

Remove secrets before creating the test

Runbooks and logs often contain more sensitive information than their owners realize: internal hostnames, user identifiers, tokens copied during troubleshooting or customer references.

For a first pilot, create a sanitized copy with synthetic values. Do not use an AI pilot as the moment you discover that your incident export contains credentials.

Apply the same classification and retention rules you already use for the underlying content. File format does not change the sensitivity of the information.

Use the right tool for the right knowledge

Notebooks are a good fit when a user wants to work with a focused set of references for a project or analysis. If you need an organization-wide authoritative knowledge source with a formal publishing process, design that separately.

The fact that Markdown now works is valuable because technical teams no longer have to convert every useful text artifact before using it in a notebook.

Final thoughts

This is a small feature with very practical potential. Start with a clean runbook and a synthetic log, ask a question that requires evidence, and check the citations. That will tell you much more than uploading a giant folder and asking Copilot to “analyze everything.”