Run a multi-source competitive teardown with the Researcher agent

Step 1: Frame a specific research questionCopilot

Write the question as something with a clear, checkable answer — "how has their pricing model shifted in the last 12 months" rather than "tell me about this competitor." The Researcher agent performs noticeably better against a bounded question, since an open-ended one gives it no way to judge what's actually relevant versus interesting-but-tangential.

Step 2: Point it at your source materialOneDriveSharePoint

Give it access to the internal documents and public sources relevant to the question, so its citations are traceable back to something real.

Prompt idea:

Research how [Competitor]'s pricing model has changed over the last 12 months, and what that implies about their target customer shift. Cite every claim to its source document or public reference.

Step 3: Read the citations, not just the summaryCopilot

The value of a cited report is that you can check it — spend time verifying the two or three claims your decision hinges on most.

Step 4: Turn findings into a decision, not just a documentCopilot

A teardown report is only useful if it changes something — connect the findings explicitly to the decision it's meant to inform.

Check the work

• Open the actual source behind each claim your decision depends on, not just the citation label.
• Check the research question was answered directly, not just adjacent information summarized.
• Confirm the sources used are current — pricing and positioning change fast.

Source: M365 Copilot Connection, "The Copilot Researcher Agent Got a Serious Upgrade in 2026" (2026)