AI training for nonprofit staff: where to start

Ask your staff whether they would take part in training on AI tools and most of them will say yes. Ask whether the organization has offered any in the past year and most of them will say no. That pairing shows up constantly, and it is not a budget problem so much as a nobody-owns-it problem: AI training for nonprofit staff sits in the gap between the person who would run it and the person who would authorize it, and nothing happens.

The good news is that the first useful version of this costs almost nothing, because the material is already in the building. Your colleagues are using these tools now, some of them well, and the fastest way to raise the floor is to make what those people know available to everyone else.

Start with what people already use

Do not begin by choosing a tool to train people on. Begin by finding out what is in use.

Ask, in a way that gets an honest answer, which kinds of tool people use for the organization's work and how often. Categories are enough: a general assistant, writing and grammar help, meeting notes and transcription, translation, image and design features, data and spreadsheet help, and the AI features switched on inside software you already pay for. Add one open question about which specific tools, because that is the list your training has to be about.

This matters for a practical reason. Training built around a tool nobody uses teaches people a product; training built around the tools already in the building fixes the mistakes people are already making. Invented example: a housing nonprofit planned a session on writing prompts and discovered when it asked that the most common use in the organization by a distance was the meeting notetaker, where the real risk was not prompt quality but which meetings it was being allowed into.

If the answers you get in a staff meeting are thin, that is worth noticing rather than pushing through. People are cautious about admitting to tool use when there is no written rule about it, which is one of several reasons to write one down first, using the outline in writing a nonprofit AI policy.

Building AI training for nonprofit staff around real work

Ninety minutes, everyone in one room or one call, laptops open. Not a webinar.

Fifteen minutes: the rules. What may never go into one of these tools, which tools are approved, and what has to be checked before anything goes out. Put this first, not last. People remember the beginning, and this is the part with permanent consequences. The reasoning behind the never-goes-in list is worth spending real time on, and it is set out in what never goes into an AI tool.

Thirty minutes: two people show what they actually do. Real tasks from your organization, on screen. A grant narrative first draft. A newsletter paragraph. Cleaning up a set of meeting notes. The rule for presenters is that they show the failures too, including the sentence that came out confident and wrong.

Thirty minutes: everyone tries one task from their own job. This is the part that gets cut when the session runs long, and it is the part that works. Reading about a tool teaches nobody anything. People bring one real thing they have to do this week, try it, and compare.

Fifteen minutes: what did it get wrong. Go around and collect the errors people just found. Made-up facts, invented citations, a tone that sounds nothing like your organization, a summary that dropped the one decision the meeting actually made. This segment is what turns enthusiasm into judgment, and it is the whole point.

A staff team that leaves knowing how to check output is in a better position than one that leaves knowing twelve clever techniques.

Who runs it, and what it costs

The person on your staff who is already good at this is usually the right choice, and asking them to prepare an hour is a reasonable use of their time. It also does something for them: at a small organization there is often nowhere to be promoted to, and being the person who taught the team something is one of the few forms of growth available, which matters more for keeping people than most boards assume. That argument runs through why people leave small nonprofits.

Outside options exist and are worth a small budget line. Candid maintains a library of free guidance and training for nonprofit staff, and state associations of nonprofits increasingly run sessions on this specific topic at prices a small organization can meet. A peer organization that is a year ahead of you will often walk your team through what they learned for nothing.

What to avoid is a vendor session dressed as training. It will be competent and it will teach your team that product, which is not the same as teaching them judgment, and it will skip the section about when not to use it.

Training is not the same as adopting

This is worth saying out loud at the start of the session, because otherwise half the room hears the training as an announcement.

An organization can train its staff and still hold a cautious position, and the two are not in tension. If your position is to hold off, the training people need is about what the position is, what got switched off, and what to do when a partner sends them something drafted by a tool. If your position is to try a few things carefully, the training is about those things and the limits around them. In both cases the honest session includes the tasks where these tools are a bad idea, and a trainer who never mentions any is selling something.

Repeat something short every few months rather than running one big session a year. Twenty minutes at a staff meeting, one person showing one thing that worked and one that did not, is enough to keep the practice from drifting.

If you want to know what your team would actually use before you plan any of it, the AI readiness assessment asks staff how often they use these tools, which kinds, how they rate their own skill, and whether they would take part in training. A staff section is released only at 4 or more responses covering at least 60% of the roster, and nothing in the report can be traced back to a person. It is free and takes about seven minutes to answer.

Most people at your organization are learning this in the ten minutes between two other tasks, alone, from whatever they found first. Ninety minutes together, built around the work they already do, replaces a lot of that.