Copilot in Excel Filtered the Table? Check the Visible Rows Before You Summarize

A filtered spreadsheet can make a correct summary look wrong. The calculation may be fine. The table you gave it may simply be showing only part of the data.

This guide uses Microsoft Copilot in Excel to filter a fictional expense table and create a review note. The point is not to trust a neat answer. It is to prove which rows were visible when the answer was made.

Evidence label: documentation-based guide with an illustrative example. Microsoft documents that Copilot in Excel can work with formatted table data and apply sorting and filtering. The sample workbook and expected output below are editorial examples. No Copilot run or personal test is claimed.

What you will create

You will make a small filter audit containing:

the filter rule
the row IDs visible after filtering
the rows that were hidden
the summary question you can answer from the filtered view
a decision to use, clear, or redo the filter

This is for an office professional checking an expense summary before sharing it with a manager.

Before you start

You need Excel for Microsoft 365 with Copilot in Excel available on your account or workspace, plus a blank workbook or an approved copy. Microsoft says you should select a cell inside the table or data range before using Copilot. Availability and controls vary by account.

Do not use payroll, customer, card, or other sensitive data for this practice run.

Fictional input

Create a sheet named October expenses and paste this table into cell A1:

Expense ID,Team,Category,Amount EUR,Status
E-101,Sales,Travel,420,Approved
E-102,Sales,Software,95,Approved
E-103,Support,Travel,180,Pending
E-104,Support,Software,240,Approved
E-105,Operations,Travel,610,Pending
E-106,Operations,Software,75,Approved
E-107,Sales,Travel,130,Rejected
E-108,Support,Travel,320,Approved

The practice question is: “Which approved travel expenses are above 300 EUR?”

From the source table, the expected row IDs are E-101 and E-108. That is an editorial check against the fictional input, not a Copilot result.

Three steps

1. Make the source boundary visible

Select the full range, including the header row.
Format it as an Excel table and name it OctoberExpenses.
Check that all eight Expense IDs appear once.
Select a cell inside the table before opening Copilot.

Keep the original table unchanged. If Excel shows a filter icon in a header, record that a filter may already be active.

2. Ask Copilot for a bounded filter

Open Copilot in Excel and use this prompt:

Use only the OctoberExpenses table.

Filter the table to show rows where Category is exactly Travel and Status is exactly Approved.
Do not delete, overwrite, or rename any source rows.
Return a review note with:
1. the exact filter conditions,
2. every visible Expense ID,
3. the Amount EUR for each visible row,
4. the total of the visible Amount EUR values,
5. a warning that the result covers only the filtered rows.
If the table, headers, or existing filters are unclear, stop and say what is missing.

If Copilot offers to change the workbook, work on a copy. Microsoft documents filtering support within a single table, but the exact buttons and chat controls can vary by surface.

3. Clear the filter and compare the row IDs

Copy the visible Expense IDs into your audit note.
Clear the table filter using Excel’s filter control, or use the ordinary Excel clear-filter command.
Confirm that all eight source rows are visible again.
Compare the filtered IDs with the source table. For this fictional input, the expected set is E-101 and E-108.
Add the following note:

Filter: Category = Travel AND Status = Approved
Visible IDs: [copy the IDs]
Visible total: [copy the amount total]
Source row count: 8
Filtered row count: [count the visible IDs]
Decision: [USE FOR FILTERED QUESTION / CLEAR AND RECHECK / DO NOT USE]
Human check: [what still needs confirmation]

The expected visible total for the sample is 740 EUR, calculated as 420 + 320. This is a hand-checked result from the fictional input, not a generated result.

Acceptance checks

The audit passes when:

the filter uses both conditions, not either condition
the visible IDs are exactly E-101 and E-108
each visible amount matches the original row
the hand-checked total is 740 EUR
all eight rows return after the filter is cleared
the summary says it covers approved travel expenses only
no source row was deleted or overwritten

A matching total is not enough. The row IDs are part of the answer.

Troubleshooting

Rows are missing after you clear the filter. Check for a second active filter, hidden rows, or a manually hidden range. Compare the visible IDs with the original eight IDs before using any total.

Copilot returns software expenses too. Restate both conditions with the words “exactly” and “AND”. Then inspect every returned Category and Status value yourself.

Copilot cannot see the table. Confirm that your cursor is inside a formatted table, that the headers are clear, and that the workbook is open in a supported Excel surface. If access still fails, use Excel’s ordinary filter controls and keep the same row-ID audit.

The result includes a chart or summary but not the source rows. Ask for Expense IDs and original values. Do not accept a chart as the audit trail.

Limits and decision

This check shows whether a filtered view matches a stated question. It does not approve an expense, confirm that a category is correct, or replace a finance policy. Keep a person responsible for the data, the filter definition, and the final decision.

TRY this on a practice table or approved copy when the question has clear filter conditions.

SKIP Copilot when you cannot identify the source range, active filters, or account surface. Use ordinary Excel filtering and the row-ID note instead.

USE the filtered summary only after the IDs, values, and filter conditions match the source.

For a related Copilot review of unusual rows, see Use Copilot in Excel to Find Outliers Before You Write the Summary. Then use the checked table to write the business summary, not the other way around.

Want one practical AI-for-work lesson each week? Join The Weekly Verdict.

Sources

Evidence

Documentation-based troubleshooting guide with complete fictional input and illustrative expected output. No Copilot execution, screenshot, measured result, or personal test is claimed.