How to analyse a spreadsheet with AI

By the end of this, you will have clear, plain-English insights drawn from your spreadsheet, along with a way to check those insights are actually supported by the numbers.

What you will need

• Any AI chat tool: Claude, ChatGPT, or Gemini all work well.
• Your spreadsheet data, ready to copy and paste as text, or a file you can upload if your tool supports it.
• About 20 to 30 minutes.

Step 1: Show the AI your data and explain it

Paste in your data, then explain what each column means. Numbers on their own can be ambiguous, so a quick description helps the AI read them the way you do.

I'm going to share some spreadsheet data. Here's what the columns mean:

- [Column A]: [what it is]
- [Column B]: [what it is]

Please read it and tell me back, in one sentence, what you think this data is
about. Here it is:

[paste your rows, including the header row]

What to expect: the AI will describe your data back to you. If its description is off, correct it now, before you ask any real questions.

If it does not work: if the columns come out jumbled, paste a smaller sample first and confirm the AI reads it correctly before adding the rest.

Step 2: Ask one clear question at a time

Resist the urge to ask everything at once. A single, specific question gets a far more reliable answer than a broad “tell me about this data.”

Based on this data, which three [products / months / regions] had the highest
[sales], and what was the total for each? Show the figures you used so I can
follow your working.

What to expect: a direct answer with the supporting numbers laid out. Asking it to show its working makes the answer far easier to trust and to check.

If it does not work: if the maths looks wrong, say “Please recalculate step by step and show each total.” Walking it through slowly usually fixes arithmetic slips.

Step 3: Look for patterns and explanations

Once the basic facts are clear, ask the AI to help you interpret them. This is where a spreadsheet turns into a story you can act on.

Looking at this data as a whole, what are the two or three most interesting
patterns or changes? For each one, tell me what in the data makes you say so,
and flag anything that might just be coincidence.

What to expect: a short set of observations, each tied back to the numbers. Treat these as leads to investigate, not final conclusions, and keep the ones that match what you already know about your situation.

If it does not work: if the patterns feel generic, give more context: tell it what changed in the business over the period, and ask it to connect the data to those events.

Step 4: Verify before you rely on it

This step matters, so do not skip it. AI can occasionally state a figure with full confidence that does not match your data, so spot-check the key numbers.

List every specific number you've given me in this conversation. For each one,
tell me exactly which rows of the data it came from, so I can verify it.

What to expect: a checklist tying each figure back to your rows. Pick two or three of the most important numbers and confirm them yourself in the original spreadsheet.

Wrapping up

You now have insights you can trust, because you have traced the important numbers back to the source rather than taking them on faith. From here you can ask the AI to turn the findings into a short summary, a set of talking points, or the outline of a report.

A note on big spreadsheets: an AI tool can only hold so much text at once, a limit called the context window. If your data is very large and the AI seems to lose track of the early rows, work with one section, or one summary table, at a time rather than pasting everything in one go.