Almost every article about AI for nonprofits is a list of twelve tools, which is the least useful thing you could hand a three-person organisation. A small nonprofit does not have a tool gap. It has a capacity gap, and the question is which of the jobs nobody has time for could be made smaller without putting donor trust at risk. That is a different starting point and it produces a different answer.
Start From The Bottleneck, Not The Tool
Write down the five tasks that routinely slip past their deadline at your organisation. For most small nonprofits that list looks something like this. The newsletter that goes out a fortnight late. The grant report drafted the night before. Thank you letters to small donors. Social posts for an event already half over. Minutes from a board meeting six weeks ago.
Now mark each one with two things. How many hours it takes, and whether a wrong first draft would be caught before anyone outside sees it. The tasks with high hours and a safe review step are where to begin. Everything else waits.
In practice that second test matters more than the first. A drafting tool that produces something nobody checks is a risk. One that produces something your director reads before it goes out is a time saving with a safety net built in.
The Four Jobs That Usually Qualify
First drafts of the things you already know. Newsletter copy, event descriptions, social posts, thank you letters. You know what happened. The hours go on starting, not on knowing. Our guide to digital marketing for nonprofits covers what these pieces need to contain.
Reformatting between audiences. One programme update becomes a board summary, a funder paragraph, a social post and a line for the newsletter. This is repackaging rather than invention, which is the mode these tools are reliable in.
Reading long documents you have to read anyway. A thirty-page funding guideline, summarised into eligibility, deadline and what they want to see, before you read it properly yourself. Summarise then verify, never summarise instead of reading.
Meeting notes into minutes. Tedious, time consuming, and reviewed by the board before approval, which is the safety step already built into your governance.
What Ai For Nonprofits Should Not Touch
Three categories, and the first is the one organisations get wrong.
Anything describing beneficiaries. A tool asked to write warmly about the people you serve will produce sentiment that is not yours and detail that is not true. Service user stories are earned, consented to and written by someone who was there. This is an ethics line before it is a quality one.
Figures, outcomes and impact claims. Numbers in a grant report or an annual review come from your records, checked. A plausible number in a funder document is the worst failure available to you, and these tools produce plausible numbers readily.
Anything a funder will treat as your own analysis. Some funders now ask about AI use in applications. Answer honestly. Using assistance to draft and then writing the substance yourself is a defensible position. Submitting generated analysis is not.
The Disclosure Question, Answered
Boards ask this early and it deserves a straight answer rather than a policy document.
Tell your board you are using it, in which tasks, and who reviews the output. Put it in writing on one page. If a funder asks, say so plainly. You do not need to label a newsletter, any more than you label which word processor you used, because the organisation stands behind the published text either way.
Where disclosure matters is anything presented as testimony, analysis or lived experience. The line is whether a reader would feel misled to learn how it was produced. Our guide to public relations for nonprofits covers the trust side of how you communicate.
A First Month That Does Not Overwhelm Anyone
Week one, pick the single task from your slipping list with the most hours and a safe review step. Do it with assistance once, by hand, yourself.
Week two, do it again and note whether it was faster the second time. Most tasks are, because you learn what to supply up front.
Week three, write the one page note for your board. What you are using it for, what you are not, who reviews output.
Week four, decide whether to add a second task or stop. Stopping after one is a legitimate outcome and better than half adopting five.
Notice there is no procurement in that month. Free tiers of the general assistants cover all four qualifying jobs, which matters when the alternative is a subscription competing with programme spend. Our guide to nonprofit PR on a budget covers the same discipline applied to communications generally.
The organisations getting value here are not the ones that adopted most. They are the ones that picked the single worst task and made it smaller. Our piece on media relations for nonprofits covers where those recovered hours are best spent.
If you would rather have the communications side supported by people who work with nonprofits, that is what our public relations service covers. You can find us, and how we work, on AmpLocal’s Google Business Profile.
Frequently Asked Questions
Do we need board approval before using AI tools?
Not usually for drafting on free tiers, but tell them anyway. A one page note covering what you use it for and who reviews output prevents the conversation happening badly later.
Can we use it to write grant applications?
Use it for structure and first drafts of sections you already know. The substance, the figures and the analysis must be yours, and some funders now ask directly about AI use.
Is it safe to put donor data into these tools?
Treat anything you paste as leaving your control. Never include names, contact details, donation amounts or anything about service users.
Will our donors mind?
Donors care about where their money went and whether your reporting is honest. Drafting assistance on a newsletter does not touch either. Generated impact claims touch both.
What should a three-person team do first?
Pick whichever task on your slipping list eats the most hours and gets read by someone before it leaves the building. Do that one, once, and judge from there.
