Which tool to choose
To fix a spreadsheet, the AI assistant does two different jobs.
- Errors in the formulas (the codes like #REF! or #VALUE!): you paste the formula to the assistant and it explains the error and the fix. It works with Excel and Google Sheets the same way.
- Errors in the data (dates in different formats, numbers treated as text, duplicates, hidden spaces): ask the AI for a checking formula or a macro that spots them in bulk.
- Inside Google Sheets you can also use the built-in AI assistant, which generates formulas and fixes from a request in plain language directly in the sheet.
How to do it
From Excel or Google Sheets, the method is the same.
Identify the error. Does a cell show a strange code? Is a total clearly wrong? Doesn't a column add up? Note where and what.
Bring the error to the AI. For a broken formula, copy and paste it with the error message. For a suspect piece of data, describe what you see and what it should be.
The operating syntax, when you ask it to explain and fix a formula:
This formula in Google Sheets gives me a #REF! error: =SUM(B2:B10)*C1 I deleted and recreated column C. Explain to me why it gives an error and give me the correct formula to multiply the sum of B2:B10 by the value in the new discount cell.Apply the fix and verify. Paste the corrected formula and check that the result is the expected one on a case you know.
For errors in the data, ask for a bulk check. If the problem is widespread (a hundred dates in the wrong format), ask the AI for a formula that converts them all or a macro that cleans up the column.
The operating syntax for the data:
In column A I have numbers that the sheet treats as text (left-aligned and they don't add up). Give me a formula in a column alongside that converts them into real numbers, and explain to me how to then turn that column into fixed values.Keep the original. Before applying bulk fixes, duplicate the sheet. A wrong fix across a whole column is worse than the original error.
A concrete example
Marco receives a sheet where the amounts column doesn't add up: the numbers are actually text. The sum gives zero. He pastes the situation to the AI with the instruction from the example. The assistant explains that the numbers are stored as text and gives him a formula that converts them in a column alongside. Marco applies it, now the sum works, then he turns the column into fixed values as the AI showed him. An error that had been blocking him for half an hour solved in five minutes, with the explanation of why it had happened.
When it does NOT work (and how to fix it)
If the AI fixes the formula but the result stays wrong
The error wasn't in the formula, it was in the data feeding the formula. Fix: ask the AI to help you check the source data too ("the formula is right but the total is wrong, what do I check in the data?"). Often the culprit is a cell with a space, a number as text or a hidden row.
If you can't describe the error to the AI
You see a code like #N/A and don't know what to say. Fix: copy the error code and the formula that generates it exactly, and it's enough to write "explain to me what this error means and how to fix it". The AI recognizes the standard spreadsheet codes and starts from there.
If the bulk fix ruins the good data
A formula applied to the whole column also touches the cells that were fine. Fix: always work on a copy of the sheet, and ask the AI for a fix that acts "only on the cells that have the problem", not on all of them indiscriminately. Check a sample before extending.
If the dates won't be corrected
Dates are the most stubborn source of error in spreadsheets, because a sheet can read them in different ways (day/month versus month/day). Fix: explain to the AI the exact format you have and the one you want, with an example ("I have 03/04/2026 which I mean as April 3, but the sheet reads it as March 4"), and have it give you the correct conversion.
A tip from someone who actually uses it
Don't just make the error disappear: have it explain why it happened. The difference between those who fight the same errors for years and those who stop making them is just one thing: the latter understood the cause. When the AI fixes a formula for you, also ask it "why was it wrong?". That answer is worth more than the fix, because next time you avoid the error on your own. The tool solves the symptom; understanding the cause cures the problem.
Frequently asked questions
What do codes like #REF! or #VALUE! mean?
They're the sheet's way of telling you what kind of error there is: #REF! usually means that a cell the formula refers to was deleted, #DIV/0! that you're dividing by zero, #VALUE! that you're using text where a number is needed. Paste the code to the AI and it explains it in your case.
Can the AI check a whole sheet looking for errors?
It doesn't look at your sheet from the outside, but if you describe the structure to it or paste it the suspect formulas, it helps you find the problems. For a systematic check, ask it for a formula or a macro that flags the anomalous cells (empty where they shouldn't be, errors, duplicates).
Does it work the same on Excel and Google Sheets?
The logic yes, the syntax of some formulas changes slightly between the two. That's why it's worth telling the AI which of the two you use: it'll give you the formula in the right version, avoiding a function that's correct for one not existing in the other.
If the AI gives me a complicated formula, should I trust it?
Trust, but verify. A long formula can be correct and still needs to be tested on cases whose result you know, because an error in a sheet propagates silently: a wrong total throws no warnings, it just looks like a number. Before basing a decision on a calculation, check it on two or three rows by hand. Blind trust in a spreadsheet is the number one cause of the errors nobody realizes they have.