Deciding not to use AI, and recording it well

You asked your board and your staff where the organization should be with AI tools over the next two years, and in both groups the largest answer was to hold off. That is a clear result, it came from the people who do the work and the people who govern it, and deciding not to use AI on that basis is a decision the organization should write down and stand behind.

It is worth saying plainly, because most writing on this subject treats holding off as a stage rather than a choice. A board that decides not to use these tools has done its job: it considered a question, took a view, and set a boundary for the organization. That is the same work as deciding to adopt something, and it deserves the same paperwork.

Deciding not to use AI is a real position

The reasons organizations land here are usually good ones, and they are worth writing into the record rather than leaving as a general unease.

Obligations you already carry. If you hold health information, education records, immigration case files, or data under a government contract, you have commitments about disclosure that were made long before any of this existed. An organization that decides it cannot meet those commitments while these tools are in the workflow has reasoned correctly, not timidly.

What the people you serve would expect. A domestic violence agency and a community theatre are not in the same position on this, and neither is wrong. If your relationship with participants rests on absolute discretion, the calculation is different, and the board is the right body to say so.

Capacity. Every tool has a cost in attention: choosing it, learning it, checking its output, and finding out what it did wrong. A five-person organization that concludes it does not have that attention to spend this year has made a resource decision, which is squarely a board's business.

Judgment about the work. Some organizations conclude that the writing, the listening, and the relationships are the work, not the overhead around it, and that a faster draft is not worth what it changes. That is a view about mission, and a board is allowed to hold it.

None of those requires anybody to be against technology in general, and a board should resist the framing that it does. Invented example: a legal aid nonprofit adopted a hold-off position with a single sentence of reasoning about client confidentiality under their funding agreement, and it was the least contested vote of the year, precisely because it was a specific reason rather than a mood.

Making it true in practice

Here is the part that catches organizations out, and it has nothing to do with willpower.

You are probably using some of these tools already, because they arrived inside software you already pay for. The notetaker in your meeting platform. The summarize button in the inbox. The writing suggestions in your document editor. A feature added to your donor database in a release nobody read. A position that says we do not use AI, adopted in a building where four of those are switched on, is untrue the day it passes, and everyone who notices learns to treat the position as decorative.

So a hold-off position needs an inventory, and it needs one more than an adoption position does.

List what is on. Ask the person who administers your accounts to go through your main systems and write down every AI feature that exists and whether it is enabled. This usually takes an afternoon and is regularly surprising.

Decide each one on purpose. Some you will switch off. Some you will keep, because a spelling suggestion is not what anyone meant. What matters is that the list is decided rather than inherited.

Say what happens at the edges. Can a staff member use one of these tools for personal work on a personal account. What about a contractor drafting your grant application, or a consultant who uses one to write the report you commissioned. What do you do when a funder sends you a form asking about it. Write a sentence for each rather than discovering your answer in the moment.

Keep the never-goes-in rule anyway. Even an organization that uses none of these tools should have written down which information may not go into one, because the rule is what protects you when somebody tries something without asking. The reasoning is in what never goes into an AI tool.

Recording it, and naming a date

The minute needs five things, which are the same five any position needs.

The position in one sentence. The reasoning, briefly, because the reasoning is what a future board will need and what a funder may ask for. What it means in practice, meaning the inventory decisions above. Who to ask when a situation is not obviously covered. And a date to look at it again.

Six or twelve months is normal. Put it on the annual calendar the way you schedule the audit or the insurance renewal. A review date is not a concession or an admission that the position is provisional. An adoption decision would get exactly the same treatment, for the same reason: circumstances change, contracts change, and a position nobody ever revisits stops being a decision and becomes a habit.

Write down what would prompt an earlier look, too. A change in your funding terms. A tool you rely on making the feature unavoidable. A specific piece of work becoming impossible without it. Naming the triggers in advance is what stops the next conversation from being a fight about whether to have the conversation.

Telling staff what the position is

Say it once, clearly, in words nobody has to interpret. Something like: we have decided not to use AI tools for the organization's work at this time, here is what that covers, here is what we switched off, here is who to ask, and we will look at this again in June.

Three things to get right in the telling.

Do not deliver it as a reprimand. If people were using these tools before the decision, they were doing so in the absence of a rule, which is an organizational gap and not a personal failing. A hold-off position announced as a crackdown produces quiet use rather than no use, and you will have lost the ability to find out what is actually happening.

Give a route for the awkward cases. Somebody will get a document from a partner that was obviously drafted with one of these tools. Somebody will find the transcription feature switched back on after an update. A named person to ask keeps those from being resolved privately.

Say it more than once. A position announced in one all-staff email is invisible to everyone who joins later, which is the most common way a written position stops being known. The fixes are the same regardless of which direction the position runs, and they are in the AI policy nobody knows about.

For general guidance on adopting and documenting policies at a small organization, including templates you can start from, the National Council of Nonprofits is the usual first stop, and your state association will know what peer organizations in your field are doing.

If your two groups did not agree, or if the direction answers were split rather than clear, that is a different situation and it needs a different meeting, described in how a board reaches a position on AI.

If you have not asked the question yet, the AI readiness assessment asks board members and staff where they think the organization should go, and reports a majority for holding off as a position to record rather than a gap to close. It is free, takes about seven minutes per person, and no answer in it can be traced to the person who gave it.

A board that says no, in writing, with reasons and a date, has governed. The organizations that end up in difficulty are not the ones that decided to hold off. They are the ones that never decided anything and found out what their position was afterwards.