28. Copilot and AI in Excel

28.5 The COPILOT Function: Retired in September 2026, and What to Use Now

By Ravindra Bagale · Last updated: 7 October 2026

Example first. Salman opens last year's feedback workbook. Column C used =COPILOT("Classify",B2,"as Delivery, Quality, Price or Other") and still shows the old categories. He corrects one note, the cell recalculates, and it now shows #NAME?. Microsoft retired the COPILOT worksheet function on 14 September 2026. This lesson explains what the function did, what happens to old files, and the two replacements: the Copilot pane, and a plain formula when the rule is clear.

First look at the data (before → after):

Expected result — Rule-based formula (same answer every time)

Replace the formula with a rule list:
IFS(REGEXTEST(B2,"\blate\b|delay",1),"Delivery", … ,TRUE,"Other")

What the function was (for reading old files)

Syntax: =COPILOT(prompt_part1, [context1], [prompt_part2], [context2], ...). The prompt parts were text; the context arguments were cells or ranges whose values were sent with the prompt. Microsoft's example: =COPILOT("Classify", B1:B10, "into one of the following categories ", A1:A4). Microsoft's own limits while it existed: about 100 COPILOT calculations every 10 minutes, no calculation in workbooks labelled Confidential or Highly Confidential, internet required, and it was not meant for exact numeric calculations.

Microsoft's replacement table (from the retirement notice)

Better still: a formula, when the rule is clear

An AI answer can change from one run to the next; a formula cannot. When you can write the rule down, write it as a formula. Here are 8 fictional feedback notes and an ordered rule list using REGEXTEST (lesson 9.12). 1 as the third argument means "ignore upper and lower case".

C2: =IFS(REGEXTEST(B2,"\blate\b|delay",1),"Delivery",REGEXTEST(B2,"damaged|stale|warm|melted",1),"Quality",REGEXTEST(B2,"price|expensive|costly",1),"Price",TRUE,"Other")

Look at the last row. With the first rule list, "Chocolate bar melted" becomes Delivery, because "Chocolate" contains the letters late. The fix is the word boundary \b (so only the whole word late matches) and adding melted to the Quality words. This is exactly the kind of mistake you only catch by reading every row, whether a formula or an AI did the classifying.

Category counts with the fixed rules: Delivery 2, Quality 3, Price 2, Other 1 (total 8). Check: =COUNTIF(C2:C9,"Quality") returns 3.

When to use the Copilot pane instead. Free text that no keyword rule can handle ("the rider was polite but the bag smelled of onions"), summaries, and translation. Prompt: "Classify each note in B2:B9 as Delivery, Quality, Price or Other and add the result in a new column called AI Category. Do not change column B." Then compare the AI column with your rule column: =SUM(--(C2:C9=D2:D9)) counts the rows where they agree, and you read the rows where they do not.

Microsoft sources (checked 7 October 2026): COPILOT function (retired); Get started with Copilot in Excel.

Practice task (try it before you open the answer):

1. Classify the new note "Delivered late, and the price was high" with the fixed rule list.
2. Say why it is not classified as Price.
3. Change the formula so Price is checked first and say the new result.