You open a spreadsheet. Hundreds of rows, a dozen columns, and a deadline. Somewhere in there is an answer your manager needs — a trend, a total, a red flag — but finding it used to mean knowing which formula to write, and getting it exactly right.
That's changed. Microsoft has built AI directly into Excel through Copilot in Excel, and it now handles a surprising amount of real analysis: summarizing data, building formulas, cleaning up errors, and even running statistical models — all from a plain-English question. Here's specifically what it does, how to turn it on, and what to double-check before you trust its output.
Copilot in Excel (chat panel). A side panel where you type questions about your data in plain language and get back summaries, PivotTables, charts, and trend callouts.
Formula generation. Describe what you want — "pull unique customer names sorted by total spend" — and Copilot writes the formula for you, using functions like FILTER, SORT, UNIQUE, or XLOOKUP.
The =COPILOT() function. A newer addition that lets you embed an AI-generated answer directly inside a cell, so it updates like any other formula.
Copilot with Python. For heavier analysis — forecasting, clustering, correlation, statistical testing — Copilot can write and run Python code (using pandas, matplotlib, and scikit-learn) behind the scenes and hand you back a chart or table. You never touch code yourself.
Copilot in Excel requires a Microsoft 365 Copilot license — it isn't included in every Microsoft 365 plan. It also works best, sometimes only, when your data is formatted as a proper Excel Table rather than loose rows and columns.
- Select your data range and go to Insert > Table so Excel recognizes it as structured data. This single step fixes more Copilot accuracy problems than anything else.
- Click anywhere inside the table, then find the Copilot icon on the Home tab of the ribbon.
- A panel opens on the right. Type your question in plain language — or use voice — and press enter.
- Copilot reads your table, works out what you're asking, and returns an answer as text, a chart, a PivotTable, or a formula, depending on what you asked for.
That's the whole setup. No add-ins to configure, no macros.
Rather than "AI can summarize data" in the abstract, here are prompts you can copy directly into the Copilot panel:
- "What's the fastest-growing category in this data, and by how much?" — Copilot scans the table and returns a direct answer plus the numbers behind it.
- "Show me total sales by region as a chart." — generates a chart without you touching Insert > Chart.
- "Give me a formula that pulls the top 5 customers by revenue." — Copilot writes a SORT/FILTER formula and explains what it does.
- "Find duplicate rows and anything that looks like an error." — flags inconsistencies you'd otherwise catch by manually scrolling.
- "Forecast next quarter's revenue based on this trend and show me the confidence interval." — this is a Python-powered request; Copilot generates and runs the statistical model for you and returns a chart.
Say you've got a year of expense data and your manager asks, "Where are we overspending?" The manual way: build a PivotTable, sort by category, compare to budget line by line, format a chart, write up your findings. Easily an hour.
With Copilot in Excel: put the data in a Table, open the Copilot panel, and ask "which expense categories are over budget, and by how much?" Copilot returns the categories, the dollar amounts, and can generate a chart on request — often in under a minute. If your manager also wants a trend line for next quarter, that's a follow-up prompt asking Copilot to forecast using Python, not a separate skill you need to go learn.
AI-generated analysis is fast, but it isn't infallible, and treating it as gospel is where people get burned:
Check the formula it gives you, not just the result. Click the cell and read the formula Copilot inserted. If it doesn't make sense to you, ask it to explain — that's a valid prompt too.
Garbage in, garbage out. If your source data has inconsistent categories, mixed date formats, or blank cells, Copilot's summary will inherit those problems. A quick "clean this data" prompt before you analyze it catches a lot.
It needs a real Copilot license. If you don't see the Copilot icon in your ribbon, your organization's Microsoft 365 plan likely doesn't include it — that's an IT/licensing conversation, not a settings toggle you're missing.
Verify anything that goes into a report leadership will see. Spot-check totals against a manual filter or a known number before you present AI-generated figures as fact.
Employers increasingly expect employees to work with data, even outside finance and analytics roles. Being able to open a spreadsheet, ask a clear question, and come back with a real answer — quickly and independently — signals that you're current and not waiting on someone else to interpret numbers for you.
This is exactly the kind of skill gap WBTS was built to close. Copilot in Excel rewards curiosity and comfort asking questions, not formula memorization.
Pick one spreadsheet you already use. Turn it into a Table (Insert > Table), open the Copilot panel, and ask it to summarize the data or identify a trend you'd normally dig for manually. Then try asking it for a formula instead of building one yourself. Small, real reps like this build the confidence to use it on something that actually matters.
Get structured, hands-on practice instead of piecing it together on your own — including the AI & ChatGPT Masterclass and Mastering Microsoft Excel courses.
