Excel's October 2026 update goes after exactly these jobs. This guide explains what each new feature actually fixes, how to use it step by step, what to double-check before you trust the output, and how to get the same result if you don't have Copilot yet.
The update in one minute
There are five new Copilot features and one everyday improvement:
| Feature | The problem it solves | Where | Status (October 2026) |
|---|---|---|---|
| Clean data | Messy data that breaks formulas and PivotTables | Windows, Mac, web | Rolling out |
| Get data | Importing and shaping data with Power Query | Web | Available |
| Skill builder | Repeating the same multi-step task every month | Windows, Mac, web | Rolling out |
| Skill recorder | Same as above, by demonstration | Windows and Mac (Insiders), web | Partly Insiders |
| Interactive canvas | Rebuilding status reports by hand | Windows | Insiders only |
| Name tables at creation | Tables stuck with names like Table1, Table2 | Windows | Available |
The first five need Copilot in Excel. The last one works for everyone on Excel for Windows. Not sure if you have Copilot? See "Who can use these features?" near the end.
1. Clean data: find the errors before they break your report
The problem: Data exported from ERPs, CRMs or other people's files is rarely clean. One trailing space makes XLOOKUP return #N/A. A date stored as text lands in the wrong month in a PivotTable. "USA", "US" and "United States" show up as three separate countries. These errors are hard to spot because the numbers still look right.
What the feature does: The clean data skill scans your data for five common problems:
- inconsistent text and number formats,
- inconsistent dates,
- duplicate values, and
- values stored as the wrong data type (such as numbers stored as text).
It lists what it found, explains why each issue matters, and proposes fixes. Nothing changes until you choose which fixes to apply.
1. Format your data as a table (select it and press Ctrl+T).
2. Open Copilot and type @clean-data, or choose it from the Skills menu.
3. Say what you plan to do with the data, so Copilot focuses on what matters.
4. Review each proposed fix, then apply only the ones you agree with.
Example prompt:
@clean-data Check the Sales table before I build a PivotTable by region and month.
What to check before you trust it:
- Duplicates may be real. Two identical rows can be two genuine orders. Confirm before you delete them.
- Dates are the riskiest fix. "03/04" can mean 3 April or 4 March. Check how ambiguous dates were converted.
- Keep a copy of the raw data on a separate sheet so you can compare before and after.
No Copilot? Do it the classic way:
- Extra spaces: =TRIM(A2) removes extra spaces. =CLEAN(A2) removes hidden non-printing characters.
- Numbers or dates stored as text: Select the column, go to Data > Text to Columns, and click Finish.
- Duplicates: Data > Remove Duplicates, after checking which columns define a true duplicate.
- Inconsistent names: Use a small mapping table ("US" → "United States") with XLOOKUP.
- For data you clean every month: Use Power Query, so the same cleaning steps run again with one Refresh.
2. Get data: build a refreshable import by describing it
The problem: Many monthly reports start the same way: open last month's export, copy it in, delete the columns you don't need, filter out a few rows, and fix the headers. Power Query can automate all of this, but its editor takes time to learn, so many people keep copy-pasting.
3. Skill builder and skill recorder: stop repeating the same steps every month
4. Interactive canvas: share a report that updates itself
5. Name tables at creation: the habit that makes every workbook easier
Who can use these features?
Frequently asked questions
Conclusion
This update is less about flashy analysis and more about the boring work around it: getting data in, fixing it, repeating the same steps, and sharing the result. That's where most Excel hours go, so it's the right place to aim.
But every feature still depends on three things Copilot can't do for you: well-structured data, clear instructions, and someone who checks the output. A tidy, named table with clean headers will improve your results more than any prompt.
