Copilot in Excel gets used two very different ways: people who expect it to "just understand" their spreadsheet and build whatever they ask, and people who treat it as a fast way to write formulas and explain data they already mostly understand. The second group gets far more value out of it. Here's what it's actually built to do, and how to get it set up.
Requirements before you start
- A Microsoft 365 plan with Copilot licensed (Business Standard, Business Premium, or Enterprise, with the Copilot add-on assigned).
- Excel for the web or Excel desktop (Windows or Mac) on a recent build — Copilot in Excel rolled out gradually, so older client versions may not show it yet.
- Your data needs to be in a proper Excel Table (select your range, then
Insert > Table
Step 1: Confirm your license is active
Same process as Copilot for Outlook — assigned by an admin through admin.microsoft.com > Users > Active users > Manage product licenses. If you already have Copilot in Outlook working, the same license typically covers Excel automatically, so there's nothing extra to assign.
Step 2: Format your data as a table
This is the step people most often miss, and it's why Copilot sometimes appears to "not understand" a spreadsheet that looks perfectly normal to a human.
- Select your data range, including headers.
- Go to
Insert > Table, confirm "My table has headers" is checked, and click OK. - The Copilot icon should now be available in the ribbon when your selection is inside this table.
What Copilot in Excel is genuinely good at
- Writing formulas from a plain description. "Add a column that flags any row where the amount is over $500" produces a working formula without you needing to know the exact function names.
- Explaining an existing formula. Paste in a nested IF/VLOOKUP combination someone else wrote and ask Copilot to explain what it does — genuinely useful for inheriting someone else's spreadsheet.
- Highlighting patterns in a dataset. Ask it to identify trends, outliers, or correlations in a table, and it'll suggest things worth a closer look (though always verify rather than taking the suggestion as fact).
- Generating a quick chart from a description of what you want to visualize, without manually configuring chart settings.
Where it falls short
- Very large or complex spreadsheets. Copilot works best on a single, well-structured table. Spreadsheets with multiple interconnected sheets, complex cross-references, or messy historical formatting tend to confuse it.
- Anything requiring true step-by-step multi-stage analysis. For genuinely complex statistical work, some people get better results pasting data into ChatGPT directly and working through it conversationally, since that interface is built around extended back-and-forth in a way Excel's sidebar isn't.
- Macros and VBA. Copilot in Excel is focused on formulas and analysis, not writing macros — that's a different skillset Excel's Copilot doesn't currently target well.
Always double-check formulas Copilot writes before relying on them for anything financial or reporting-critical. It's very good at producing something that looks correct; it's not infallible at producing something that is correct, especially with edge cases like blank cells, text in number columns, or unusual date formats.
A practical first task to try
If you're setting this up for the first time, a good test case is picking an existing table you already understand well, and asking Copilot to "summarize the key trends in this data" or "find any rows that look like outliers." Because you already know the data, you'll quickly get a feel for how much you can trust its output before pointing it at something you don't have full visibility into yourself.