When the board and staff disagree about AI

The board says a donor letter should never be drafted by a machine. The staff say it is fine as long as a person reads it before it goes out. Both groups answered the same question about the same organization and gave different answers, and the first instinct is usually to work out which group is right. When the board and staff disagree about AI, that instinct wastes the most useful thing you have.

The gap is the finding. Two groups who see the same organization from different distances have told you that they are working from different pictures of the work, and that is worth more than either group's answer on its own. It is also the kind of thing that stays invisible until somebody asks both groups the same question in the same words.

What it means when the board and staff disagree about AI

Three explanations cover most gaps, and they call for different responses.

The two groups are picturing different tasks. This is the most common one by a wide margin. A board member hears donor communication and imagines the annual appeal letter, which carries the organization's voice and gets read by people who have given for twenty years. A development associate hears the same phrase and thinks about 340 gift acknowledgements that have to go out this month, each one a variation on the same four sentences. They are not disagreeing about principle. They are answering about different work.

The two groups are carrying different risk. Staff bear the operational consequence of a rule that makes the job impossible, and they know exactly how many hours the current process costs. Board members bear the consequence of a story in the local paper. Both are real, and neither group can see the other's clearly from where they sit.

One group has information the other does not. Staff usually know what the tools do now, because they use them. Board members usually know what a funder said in a meeting, or what a peer organization got wrong last year. Invented example: a literacy nonprofit's board was firmly against using AI for anything touching participants until a staff member explained that the translation feature the board was worried about was already inside the video platform they had used for two years.

None of these is a discipline problem, and none of them is resolved by a vote at the end of a twenty-minute discussion.

Put one use case on the agenda, not the topic

The mistake is scheduling a discussion about AI. That produces forty minutes of general position-taking and no decision, because there is nothing specific enough to decide.

Take the single use case with the widest gap and put that on the agenda by name. Thirty minutes, one topic, and a written outcome. Something like: should we use AI tools to draft messages to donors.

Here is a structure that works.

Five minutes: what the work actually is. Somebody who does the work describes it. How many, how often, how long it takes now, what the current process is. No opinions yet. Most of the gap tends to close in this five minutes, which is why it goes first.

Five minutes: what people said. Read out the two distributions without commentary. Not who said what, which nobody knows, but the shape: most of the board said one thing, most of the staff said another, and here is roughly how many.

Ten minutes: what could go wrong, from both directions. A donor recognizing a form letter is a risk. So is the development director spending two evenings a month on acknowledgements instead of on donor relationships. Write both lists on the same board.

Ten minutes: decide the boundary, not the tool. The board's job is to say where the line is. Which uses need a person checking, which are off the table entirely, and what would have to be true for that to change. Which specific product gets used, and how, is management's call and does not belong in a board minute.

That last split does more work than it looks like. Most disagreements between a board and a staff team about AI are actually disagreements about who decides, and naming the boundary explicitly settles two things at once. The same discipline that keeps a board meeting from becoming a staff meeting applies here, and it is easier if the agenda is built around decisions in the first place, which is the argument in building an agenda around decisions rather than reports.

What to write down

One paragraph in the minutes, in this shape: for this use case, here is what is allowed, here is what is not, here is who decides the details, and here is when we look at it again.

Also write down the disagreement itself. Not as a complaint, as a record. A year from now, the useful question is whether the gap narrowed, and you can only answer that if you wrote down that it existed. Boards routinely lose this kind of information because it feels awkward to minute a difference of opinion, and then they have the same conversation from the beginning at the next meeting.

If the gap turns out to be mostly about what the tools are and what they do, the honest answer is that the board needs an hour of neutral grounding before it decides anything, which is a normal and unembarrassing conclusion. That hour has its own shape, described in what an hour of board education on AI should contain.

When the disagreement is wider than one use case

Sometimes the gap is not on one row. The board wants to hold off across the board and staff are already using these tools weekly, or the reverse: a board that wants to move and a staff team that does not want the tools anywhere near their work.

That is a direction question rather than a use case question, and it takes more than one meeting. It also needs to end in a written position rather than a mood, which is a different exercise, laid out in how a board actually reaches a position on AI.

What matters either way is that lower comfort in one group is reported and treated as lower comfort, not as resistance to be managed. A board that reads a staff team's caution as an obstacle has already decided the answer and is now doing consultation theatre, and people can tell. BoardSource has written for years about the relationship between a board and a staff team as a matter of mutual information rather than of control, and this is a live example of it.

If you want the gap measured rather than guessed at, the AI readiness assessment asks board members and staff the same questions and puts the two sets of answers side by side, use case by use case. Both sections have to clear a threshold before anything is shown, 4 or more responses and at least 60% of the roster for each group, and nothing in it can be traced back to a person. It is free and takes about seven minutes to answer.

A board and a staff team who agree about everything are usually not telling each other much. The gap is where the useful conversation is, and the only real mistake is treating it as a problem with the other group.