Lesson 4 of 6
Donor research & data
7 min read
"When did Maria last give, and how much?" AI answers instantly and sounds sure. But did it read your records — or just make something up?
Guessing from memory vs. reading the record
Ask AI a donor question with nothing to go on and it guesses from fuzzy training memory — a plausible-sounding number that may be completely wrong. Give it the actual record — the donor's giving history pasted or connected in — and the answer snaps to the real fact, and it can even show you where it came from. That's grounding: answers built on your data, not on a hunch.
The same question gives a fuzzy guess from memory and an exact answer from the record. Grounding is the difference between made-up and true.
Handle the data with care
Grounding donor research in real data is powerful — and it means you're handling personal information. Only put donor data into a tool your organization has approved for it, share the minimum the task needs, and remember that a supporter trusted you with their details for the mission, not for a chatbot. Grounded research is worth doing; doing it carelessly with a public tool is how a nonprofit ends up with a data problem.
Ground the answer in the record, but respect the record: minimum data, approved tools, and the donor's trust kept intact.
A confident un-grounded answer about a donor is worse than no answer — acting on a wrong giving history can insult the very supporter you meant to thank.
The shape of it
- —Ungrounded, AI guesses donor facts from fuzzy memory — plausible but often wrong.
- —Grounded in your real record, the answer is exact and traceable to a source.
- —Use approved tools, share the minimum, and treat donor data as a trust.
You want AI to tell you a donor's last gift. What gives you an answer you can act on?
Continue in the app
Take the whole AI for Nonprofits course — tracked
Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.