Nine out of ten nonprofits are using AI. So why does it feel like so little has changed?
If you lead a small nonprofit or association, you have probably seen it in your own office. Someone uses ChatGPT to tighten up the newsletter. Someone else asks Claude to summarize a long report. It helps, a little. But the workload looks about the same as it did a year ago.
You are not imagining it. The 2026 Nonprofit AI Adoption Report from Virtuous and Fundraising.AI surveyed 346 nonprofits. 92% said they use AI in some way. Only 7% said it has made a major difference in what their organization can do. Most, 79%, are stuck on what the report calls an "efficiency plateau": small or moderate gains, and nothing more.
The tools are not the problem. How we use them is.
What is holding most nonprofits back
The same report points to three patterns.
AI lives with individuals, not the team. 81% of organizations use AI on an ad hoc basis without writing down what works. One person finds a great way to draft thank-you letters, and that know-how lives in their head and their personal account. When they are out sick, or move on, it goes with them.
Almost nobody has a repeatable process. Only 4% have documented, repeatable workflows, and only 18% use AI across team workflows at all. Everyone starts from a blank chat box, every time.
There are no rules. Nearly half, 47%, have no AI policy. When staff are not sure what is allowed, especially with donor or client information, the careful ones hold back and the bold ones take risks. Neither leads to results.
None of this is a failing of the people involved. It is what happens when a powerful tool shows up faster than anyone has time to plan for it.
What results actually look like
At the nonprofit I lead, we moved past experimenting by focusing on a few tasks we do over and over. Our email and social media writing went from about 10 hours a week to about 2. Board meeting notes went from an hour to about 15 minutes. A large grant proposal that used to take about 30 hours now takes about 4. A person still reviews and approves every piece.
Those gains did not come from a fancier tool. They came from setting things up once, so the whole team gets the same help every time.
Three things the 7% do differently
1. They start with rules. A short written policy tells staff what can and cannot go into AI, and who approves what goes out. It sounds like a brake. It is actually what lets people move faster, because they stop second-guessing themselves. If you do not have one yet, my free AI Use Policy Starter Template is a good place to begin.
2. They pick a few recurring tasks and set them up once. Instead of "let's use AI more," they choose two or three tasks the team does every week or month: the newsletter, the board report, donor thank-yous. For each one, they save a short set of instructions, a few examples of strong past work, and a checklist for review. Then anyone on the team gets a good first draft, not just the one person who figured it out.
3. They write it down and measure it. They note how long a task took before and after. That is how you know it is working, and it is how you make the case to your board.
A simple way to start this month
Ask your team which tasks eat the most time each week. Pick the three that come up most.
Make sure you have basic rules in place first.
For one task, gather two or three examples of your best past work, with names and private details removed. Write a short set of instructions: who it is for, how it should sound, and what to avoid.
Save the instructions and examples in your AI tool so anyone on the team can use them. Both Claude and ChatGPT let you save them in a project.
Time the task the old way once, then the new way for a month. Compare.
When it works, move on to the next task.
You do not need a big budget or a technology person. You need a few hours of planning and the habit of writing things down.
If you would like help
This is the heart of my Nonprofit AI Setup: rules first, then custom prompts and three workflows built around your team's real work, plus staff training and before-and-after time tracking. If you would rather start on your own, download the free AI Use Policy Starter Template and begin with step 2 above.
Quick answers
Is a 7% success rate a reason to wait?
No. The report suggests most organizations stall because of how AI is set up, not because it does not work for nonprofits. Starting small, with rules and one task, keeps the risk low.
What is a "workflow," in plain terms?
A saved, repeatable way of doing one task with AI: instructions, examples, and a review step. Anyone on the team can use it and get a consistent draft.
Do we need paid AI accounts?
In my experience, a team or business plan is worth it. It keeps your work in organization accounts instead of personal ones and lets you share saved instructions with your team. Check each plan's data and privacy terms against your policy.
Source: The 2026 Nonprofit AI Adoption Report, Virtuous and Fundraising.AI, 346 nonprofits surveyed in late 2025, published February 2026.