What it does

Copilot Search is AI-powered enterprise search across your Microsoft 365 content and, through connectors, third-party systems. You ask in natural language and get a synthesised answer with sources, rather than a list of blue links to sift.

Key facts

• It's grounded search — it answers the question and cites where the answer came from.
• Reach extends beyond M365 via Microsoft Graph connectors into systems like ServiceNow, Confluence and others.
• It honours existing permissions across every connected source.
• Sits within the Copilot experience and rides Copilot licensing.

When to use / skip

Use it when people can never find the document they know exists, and when knowledge is scattered across M365 plus a couple of third-party systems. Skip expecting it to search systems you haven't connected — no connector, no results.

Configuration decisions

• Which Graph connectors you deploy, and their crawl and refresh settings.
• How connector content is permission-mapped so search respects source access.
• Prioritising which repositories are worth indexing first.

Gotchas

• Connectors are real project work — setup, mapping and licensing/quota — not a checkbox.
• Search is only as good as what's indexed. Users judge the whole thing by the gaps in your connector coverage.
• Permission-trimming on external content can be fiddly; a misconfigured connector can over- or under-expose.

Consultant notes

• Scope connector work honestly. Clients hear "searches everything" and don't cost the integration effort behind it.
• Content sprawl and poor metadata undermine results more than the AI does. A findability audit pays off first.
• Start with one or two high-value repositories and prove it, rather than boiling the ocean of connectors.

---

*Worth revisiting af