In February 2026, Virtuous and Fundraising.AI published The 2026 Nonprofit AI Adoption Report, the largest survey of AI use in the nonprofit sector to date. The headline number is the one everyone is sharing: 92% of nonprofits report using AI in some capacity.
The number nobody is sharing is buried four pages in.
Of those nonprofits, 79% report small or moderate gains. Another 7% report major strategic impact. The remaining 14% report no measurable benefit at all.
Read that again. Of the 92% of nonprofits using AI today, only one in twelve says it has produced a meaningful change in how their organization operates. The rest are stuck on what the report calls the efficiency plateau: faster emails, faster grant drafts, faster meeting notes — and the same outcomes they were producing twelve months ago.
This post is about what the 7% are doing differently.
Individual use is not the same as workflow use
The clearest signal in the data is buried in the methodology section. The report distinguishes between three modes of AI use:
- Individual use: a single staff member using a consumer tool (ChatGPT, Claude, Gemini) for personal tasks.
- Team use: a group of staff using the same tools, but with no shared prompts, templates, or review processes.
- Workflow use: AI embedded into a documented, repeatable process that runs the same way regardless of who triggers it.
81% of nonprofits using AI are in mode 1. A further 15% are in mode 2. Only 4% have what the report defines as documented, repeatable workflows.
The 7% who report major strategic impact overlap heavily with that 4%. Workflow use is not a sufficient condition for impact, but it is very close to a necessary one.
What the 7% do differently
Looking at the qualitative responses in the report — and triangulating against the dozen or so engagements we have run with nonprofits in the $1M to $25M range over the past year — three patterns hold:
1. Shared workflows, not shared subscriptions
A team ChatGPT seat is not a workflow. A workflow is a documented sequence of prompts, inputs, review steps, and outputs that produces the same artifact regardless of who runs it. The 7% have written these down. They live in a shared doc, a shared Notion page, or a shared Claude project. They are version-controlled. They get updated when the model behavior changes.
The 81% have a Slack channel where people share prompts that worked once.
2. Governance before scale
The same Virtuous report shows that 47% of nonprofits have no formal AI governance policy. The 7% almost universally do. The policy doesn't have to be long — most of the ones we have seen are two pages — but it has to exist, and it has to answer the four questions every funder is starting to ask:
- What data are we comfortable putting into AI tools?
- Which tools are approved for which kinds of work?
- When do we disclose AI use externally?
- Who is accountable when something goes wrong?
Without a written answer to those four questions, scaling AI use is a liability the board doesn't yet understand.
3. Measurement that survives the budget cycle
Time saved is not a metric finance can act on. The 7% measure something downstream: campaigns shipped per quarter, donors moved up the giving ladder, hours redirected from operations into program work. They tie AI investments to organizational goals on the same timeline as every other line item, which means when the board asks whether the spend is working, there is a real answer.
The 81% point at vibes.
What the next twelve months look like
If you are reading this and you recognize your organization in the 81%, the next twelve months are going to be uncomfortable. The funders who started asking about AI capabilities in 2025 — Gates, MacArthur, Ford, a growing list of community foundations — are now asking about AI governance, AI workflows, and AI outcomes. "Yes, our team uses ChatGPT" is no longer a sufficient answer in a grant narrative.
The good news: the gap between mode 1 and mode 3 is not technical. It is operational. It is closeable in a quarter, not a year, by an organization willing to write things down.
The 7% wrote things down. That is most of what separates them.
Source: The 2026 Nonprofit AI Adoption Report, Virtuous and Fundraising.AI, February 2026.