168 nonprofits, 14 EDs, and the AI question worth asking first
Wanzhu Shi and Lauren Azevedo surveyed 168 Florida nonprofit organizations. They also sat down with 14 Florida executive directors and asked them to explain their own decisions, in their own words. The study ran in Nonprofit and Voluntary Sector Quarterly this past March, a peer-reviewed journal (journals.sagepub.com, accessed 2026-08-12). The full findings sit behind a paywall, so I can't tell you which factor mattered most. But I can tell you what they chose to measure, and that alone is worth sitting with.
They used something called the Technology-Organization-Environment framework, TOE for short. It's a standard way researchers study why any organization adopts new technology, not just AI. It looks at three things: whether the pressure to adopt comes from outside the organization, what the culture inside the organization already is, and what stance leadership takes.
Read that list again. None of it is about the software.
I've had a version of this conversation with a dozen Florida EDs this year. Almost none of them open by asking which AI tool to buy. They ask whether they're behind. Whether their board will bring it up before they're ready. Whether their program staff are already using it and just haven't mentioned it. The tool is the last thing on their mind. The ownership question comes first, every time.
Here's what that looks like once you measure it, and it isn't encouraging.
A 2026 survey by Virtuous and Fundraising.AI, fielded across 346 nonprofits, found that 92% now use AI in some capacity (virtuous.org, accessed 2026-08-12). This is a vendor-sponsored report, worth reading with that in mind, but the underlying survey numbers are real. Only 7% say AI made a major difference to their mission or their fundraising. Less than half, 47%, have any AI governance policy at all. And 81% describe their AI use as ad hoc: no documented workflow, no named owner.
Ninety-two percent adoption. Seven percent impact. That gap is the entire story, and it has almost nothing to do with which model people are using.
I've seen this exact pattern before, and it's not new to AI. A strategic plan fails the same way. A board approves it, everyone nods, and it sits in a drawer because checking whether it moved isn't anyone's job this month. Hand a program team a new tool with no owner and no monthly check, and it becomes another open tab. Hand the same tool to one person whose job is to ask, at the end of the month, "did this help, yes or no," and it becomes something the organization actually keeps.
This is what the Florida study seems to be circling, even without its numbers in hand. External pressure gets someone to try a tool once. Culture decides whether people feel safe admitting they're using it. But leadership stance, someone actually deciding this is worth owning, is the only one of the three a director controls directly, and probably the only one that turns a tool into a habit instead of a headline.
Funders are starting to notice the same gap from a different angle. Ten major foundations, including MacArthur, Ford, Omidyar, and Mellon, committed $500 million over five years this year to a pooled initiative called Humanity AI, aimed at civil-society AI governance (macfound.org, accessed 2026-08-12). MacArthur alone has already put $10 million behind it, including grants to the AI Now Institute and Brookings. That's a signal, not a mandate. But if your nonprofit chases national funding, "who owns your AI use, and how do you know if it's working" is coming to a grant application near you, probably sooner than you'd like it to.
So what do you actually do this month, with no budget and no policy department?
Don't write an AI policy. Not yet. Pick one workflow where someone on your team is already using AI, whether or not you approved it. Grant writing, first drafts of donor emails, meeting notes, it doesn't matter which. Name the one person who owns whether it's working. Not "IT." Not "everyone." One person, one workflow.
Then ask them a single question at the end of the month: did this save real time, did it actually help the work, keep it or drop it. Write the answer down somewhere the rest of the team can see it, the same place you'd post whether last month's priorities shipped. That's the whole system. No task force, no binder, no vendor call required.
It's the same discipline I ask every ED to apply to a strategic plan, aimed at something newer. Three priorities a month, not forty. One owner per priority, not a committee. A short, honest answer at month's end about whether it moved. AI doesn't need a different kind of management. It needs the same kind, applied on purpose, to something the org hasn't gotten around to owning yet.
The 168 nonprofits in that Florida study will eventually tell us which of the three factors, external pressure, culture, or leadership stance, predicts adoption best. My guess is leadership stance, because it's the only one a director can change this month without waiting on anyone else. Pressure comes from outside. Culture takes years to shift. Stance is a decision you can make on a Tuesday.
What's running on AI in your organization right now that nobody actually owns?
Comments
Post a Comment