Skip to content

Turn a dataset into a narrative brief

Move from a spreadsheet or export to a narrative brief that explains the trend, the outliers, and the action to take.

Stage: Cowork · For: End user, Manager, Champion · Level: Intermediate · Time: 20 min

When to use this

You have a spreadsheet, CSV export, or table of operational data, and stakeholders need the story behind the numbers rather than another raw chart. Cowork can inspect the data, identify patterns and outliers, draft a concise narrative, and include a supporting table you can verify.

Use this when the goal is a decision or recommendation, not exploratory analysis for its own sake.

What you'll need

  • M365 Copilot license with Cowork access
  • A spreadsheet, export, or table with enough context to explain the columns and time period
  • A decision question, audience, and any known caveats about the data

Try it now — the prompt

Give Cowork the dataset and the decision context:

Using the attached [spreadsheet/export], write a narrative brief for [audience]
answering this question: [decision question].

Please include:
1. A one-paragraph executive summary
2. The key trends in the data
3. Outliers or segments that need attention
4. A table showing the most important figures, with clear column labels
5. A recommended action and rationale
6. Data caveats, missing fields, or checks I should run before sharing

Use only the data provided unless I give you additional context. If you calculate
anything, show the formula or method in plain language.

Why this works: it frames the analysis around a decision and requires Cowork to show its method. That makes the brief easier to verify and reduces the risk of a polished story built on unclear calculations.

Step by step

  1. Prepare the data. Clean obvious duplicates, confirm column names, and include a short note explaining the time period and business context.
  2. Hand off the analysis. Attach the file, paste the prompt, and state the decision question plainly.
  3. Check the summary against the table. The story and the numbers should agree. If they do not, ask Cowork to reconcile them before revising prose.
  4. Review calculations and caveats. Make sure any percentages, rankings, or comparisons use the right denominator and time period.
  5. Pressure-test the recommendation. Ask whether the action follows from the data or needs outside context.
  6. Ask for a stakeholder-ready version: 
    Revise this into a one-page brief for [audience]. Keep the table, make the
recommendation explicit, and list the checks I should complete before sharing.

Screenshots

We deliberately don't ship screenshots that go stale — the Microsoft Copilot UI changes often. Follow the numbered steps above, which we keep current. Maintainers can regenerate fresh captures with the Playwright tool in tooling/screenshots/.

Make it better

  • Ask for alternative cuts. "Compare the story by [region / segment / product / month] and say whether the recommendation changes."
  • Create a decision slide. "Turn this brief into a single slide with headline, evidence table, and recommendation."
  • Add a caveat section. "Separate confirmed findings from hypotheses that need more data."
  • Prepare follow-up questions. "List the questions a leadership audience is likely to ask and the data needed to answer them."

Watch out for

  • Data quality still belongs to you. Cowork can spot oddities, but it cannot know whether the export is complete or current unless you tell it.
  • Outliers need context. A spike, dip, or missing value may be a real signal, a timing issue, or a data error.
  • A narrative can overfit the data. Ask Cowork to state confidence and caveats before you present a recommendation.
  • Sensitive data needs care. Remove or mask personal, customer, or confidential fields unless they are required for the task.

Where this leads (the ramp)

When the same dataset needs regular monitoring and narrative updates, the next step is an always-on agent that watches for changes and drafts the brief for review. That is Stage 5 · Autopilots.

Next: Stage 5 · Autopilots

Was this page helpful?

Found an issue or want to suggest an improvement? Open a GitHub issue  ·  Star on GitHub