Why this matters
A nonprofit board decides real things from the development report: whether to hire, whether to expand a program, whether to plan for a smaller year. If the total is off because a gift synced twice or a pledge was counted as cash, the board decides on money you don’t have.
Donor database exports are messy in predictable ways. Online giving platforms sometimes push the same gift twice, and pledges sit in the same list as gifts actually received. Spreadsheet formulas add all of it up without a word.
In this module ChatGPT checks the export before it adds anything, you decide what counts as a gift, and then it builds giving by month and by source, a chart and three takeaways for the board, each with the number behind it.
What you’ll build
A checked and cleaned gift export for Hillside Literacy Project, a made-up nonprofit: 1,212 rows in, 5 duplicate online gifts removed, 14 pledges ($18,750) marked “Pledge - not received” and kept out of every total, 1,193 gifts received.
Then two tables (gifts by month and source, $157,765 for January to June; first quarter against second quarter, up 42%), a line chart by source for the board report, and three takeaways: the Spring Appeal lifted one-time gifts 65%, monthly donors grew every month to 13% of giving, and a third of all giving came in two foundation checks.
What you need
- Nothing to start the practiceChatGPT is a practice copy and the export is made up. Nothing is uploaded.
- A ChatGPT account, to do it for realFile analysis works on every plan, with limits on Free. Use a business plan that doesn’t train on your data for donor information.
- A gift export from your donor databaseA CSV or Excel export of one period’s gifts with date, donor type, campaign, gift type and amount. Remove names, emails, addresses and card details first; a donor ID is enough.
- Your own rules for what countsHow your organization treats pledges, matching gifts, in-kind gifts and grant payments. Ask your finance lead or bookkeeper if you’re not sure.
Concepts
3 ideas you’ll use in the practice.
Check the file before you ask a question
The first prompt asks only for what each column means, the row count, the date range and anything that looks wrong, and says no insights yet. That is when ChatGPT finds the duplicates and the pledges.
If you skip it and ask for the total straight away, you get a confident number that includes both mistakes.
You decide what counts as a gift
ChatGPT can spot that 14 rows are pledges. It can’t know your policy for them. In the lesson you tell it: mark them “Pledge - not received” and leave them out of every total until they are paid.
Marking instead of deleting keeps them in the file, so you can still show the board what is pledged and not yet received.
Split the total by source
One big total hides the story. Split by source, the export shows that a third of the half year’s giving came in two foundation checks, while monthly donors grew quietly every month.
That is the kind of takeaway a board can act on: grow monthly giving, and plan for what happens if one grant comes in late or smaller.
Practice
Do every step yourself in a practice copy of the real tool. It checks your work, and “Do it for me” finishes any step you get stuck on.
Do it for real
Now repeat it in your own account. Tick each step as you go.
Downloads
- Gift report prompts (.txt)The five prompts from the lesson, in order, ready to change for your own export.
Quick check
3 questions to check you’ve got it.
1. The check finds 14 rows with the gift type Pledge. What did you tell ChatGPT to do with them?
- Count them, a pledge is a promise
- Mark them “Pledge - not received” and leave them out of every total
- Delete them from the file
Show the answer
Mark them “Pledge - not received” and leave them out of every total. A pledge isn’t money received. Marking it keeps it in the file so you can report it separately.
2. Why ask ChatGPT to check the file before asking for totals?
- It is faster
- So duplicates and pledges are found before they are added into a total the board will use
- ChatGPT can’t add up without it
Show the answer
So duplicates and pledges are found before they are added into a total the board will use. A total made from a messy export looks just as confident as a correct one.
3. Foundations gave 34.9% of everything received, in two payments. What does the double-check ask you to confirm?
- That the June payment is for this year’s grant, and that totals match recorded deposits
- That foundations like the chart
- Nothing, the numbers are final
Show the answer
That the June payment is for this year’s grant, and that totals match recorded deposits. A grant paid early for next year would make this year look bigger than it is. Your finance team knows how it is booked.
You reached the end of AI for Finance Teams
Your gift totals are checked and tie back to the deposits. Pick your next course, or reuse the prompts for your next month-end report.
Pick your next course →