Ask for formulas in plain words
You do not need formula words. Tell the AI three things: which app you use, what is in each column, and what you want to find out. Say which rows your data is in, too. Formulas can work a little differently from app to app, so naming yours avoids mix-ups.
Before
formula to add up sales
After
I use Google Sheets. Column A has dates, column B has the salesperson's name and column C has the sale amount. The data is in rows 2 to 500. Write a formula that adds up the sales for the person whose name I type in cell F1, but only for this month. Then explain each part of the formula in one plain line.
Column letters, where the data sits and the exact goal. That is what the AI needs to write a formula that fits your sheet.
If the formula shows an error, do not start over. Paste the formula and the exact error back into the same chat: "I got #N/A from this formula. What is wrong, and how do I fix it?"
Clean up messy data
Messy data is a great job for AI. Here are fixes you can ask for:
- Split "Lopez, Maria" into a first name column and a last name column.
- Make every date look the same, like 03/15/2026.
- Find rows that show up twice, based on the email column.
- Fix names typed in ALL CAPS or all lowercase.
- Put each expense in a group, like rent, supplies, travel or food.
- Trim extra spaces that stop names from matching.
For a short list, paste the rows and ask for a clean table back. For a big sheet, ask for a formula or the steps instead, so the work happens in your file and you can see every change. Either way, swap real names and contact details for fake ones first. A few made-up sample rows teach the AI your layout just as well. More on that in keep your private info safe.
Ask what your numbers say
Paste a small summary table, like sales by month or spending by group, and ask a plain question. The AI can point out what changed, what stands out and what to look at next.
You are a patient analyst who explains numbers in plain words. Here is my data: [paste a small table, with private details removed] What I want to know: [your question, like which month was best and why] Rules: - Show your math step by step so I can check it. - Use only the numbers I gave. If something is missing, say so instead of guessing. - Point out the 3 things that stand out most, like a big jump or drop. - Keep every point to one or two short sentences. End by naming one chart that would show this best, and what goes on each side of it.
AI can explain and organize your numbers, but it should not make big money choices for you. For taxes, loans or investing, take what you learn to an accountant or another licensed pro.
Pick the right chart
Not sure how to show your numbers? Ask the AI which chart fits, and for the steps to make it in your app. This table covers the usual needs:
| To show | Ask for |
|---|---|
| Change over time, like sales each month | A line chart |
| A few things side by side, like sales by store | A bar chart |
| Parts of a whole, with only a few parts | A pie chart |
| Whether two things move together, like ads and sales | A scatter chart |
| Exact numbers people need to look up | A simple table, not a chart |
Ask for a short title and a one line takeaway to go under the chart, like "Sales rose every month after we added Saturday hours." That line is what people remember.
Check the math
AI can make math mistakes, especially with long lists of numbers. These habits catch them:
- Ask it to show its work. Steps you can follow beat one number you cannot check.
- Test on a few rows. Try a new formula on 3 rows you can add up in your head.
- Let the sheet do the adding. For real totals, use a formula in your sheet, not a number the AI typed.
- Watch for numbers you never gave. If the answer has one, ask where it came from.
For a big job, like a monthly report, go one step at a time: clean the data, add it up, make the chart, then write the summary. See break big jobs into small prompts, and use how to check AI answers for mistakes for anything you will share.
Always give your column letters and the rows your data is in. Many formula mix-ups come from the AI guessing them.