Excel is where precision matters most, and where Microsoft Copilot is most useful when you treat it as an analyst you supervise rather than an oracle.
1. Prepare the data first
Microsoft Copilot works best with tidy data:
- One header row with clear, unique column names.
- One record per row, no merged cells, no subtotals mixed into the data.
- Data formatted as an Excel table where possible.
- Files saved in OneDrive or SharePoint.
Ten minutes of tidying will improve every answer that follows.
2. Questions to ask your data
Which five cost centres had the largest increase in spend compared with last quarter? Show the change in pounds and as a percentage.Are there any duplicate invoice numbers or unusual amounts in this table? Explain how you identified them.Create a pivot table of revenue by region and month, then add a chart that shows the trend.3. Formulas without the head-scratching
Add a column that shows the number of working days between Invoice Date and Paid Date.Explain what the formula in column H does, step by step, and suggest a simpler version.4. Check the result
- Recalculate a sample by hand or with your own formula.
- Ask "which rows did you include and exclude?"
- Check totals reconcile to the source system before anything goes into a report.
5. When to go further
For heavier analysis across several files, the Analyst agent can reason over data and run Python to produce charts and tables. See Researcher and Analyst.
Sources
Product details checked against these Microsoft pages in October 2026. Microsoft updates Microsoft Copilot frequently, so check the current page before relying on a detail.
Choose the right starting point for your organisation
Tell us who you want to train. Request dates for a private Copilot cohort or ask about a discovery conversation.
Prefer email? josh@igoe.ai