What changes, what doesn’t: an AI-era conversation about adoption
SignalWork Insights · May 2026
Ninety-two percent of nonprofits use AI. Seven percent see real impact.
That number is from a benchmark study Virtuous and Fundraising.AI published in February 2026, surveying 346 nonprofits in late 2025. Most teams are getting faster drafts and quicker emails. Hardly any are getting organisational transformation. The researchers called this the efficiency plateau.
I have been circling this gap for a while. A catch-up this week with a fellow Hoosier, Jay Love, gave me a chance to talk it through. Jay has built and sold four nonprofit technology companies, including Bloomerang, eTapestry, and Fund-Master. We landed in the same place: the rules have evolved, not been replaced. The playbook on internal champions still applies. What has changed are two things. First, who in the organisation has to be using the system. Second, whether the AI workflow is built into the tools the team already uses, or sits alongside them as a separate step that staff can route around.
The half that hasn’t changed: the internal champion
Jay’s framing was direct: “It all boils down to finding an internal champion. I haven’t seen that change.”
The current data backs this up bluntly. A 2025 Prosci study found that forty-three percent of AI adoption failures were attributed by respondents themselves to insufficient executive sponsorship. Not bad tools. Not unclear use cases. The absence of someone senior carrying it.
The champion question is still where this starts. It is still the single biggest determinant of whether the system gets used.
The first thing that has changed: who has to use it
Old CRM patterns let CEOs delegate. Fundraisers used the platform, development logged the calls, the CEO read the reports. The system was used by a small group of people whose job it was to use it, and the rest of the organisation worked around them.
That model is now dead. “There’s no way you can afford with so many ways of communication that it can be kept in three or four different places. It has to be in a central location.” Information now sits in Slack, email, voice notes, AI chat threads, transcripts, board portals. If the leadership team is not in the central system, the central system is not central. And in a sector where senior tenure is shrinking, “it’s never going to make it from one leadership transition to another”. Every senior departure is now a knowledge-loss event unless capture is universal.
This is the first operational shift that explains most of the seven percent. The organisations seeing real impact from AI are the ones where the CEO, the development director, the head of programmes, and the board are all in the system. Not delegated. Not summarised for. In it. The organisations on the efficiency plateau are the ones where one or two people are using AI well, and the rest of the organisation is still buried in disconnected systems while leadership reads the digests.
The second thing that has changed: closing the alternative paths
One thing worth thinking about: the most reliable way to get a workflow used is to remove every alternative path to the information it produces. If the only way to get the donor brief is through the AI workflow that generates it, the workflow gets used. If staff can still get to the same brief through three other routes, they will.
The instinct is right, and the research on adoption confirms the inverse case. The pattern that comes through consistently in 2025 and 2026 implementation studies is that AI tools stall when they sit alongside existing workflows as a separate step, and stick when they are built into the tools and templates the team already uses. The point is not really to block alternative paths. It is to make the AI-enabled workflow the path of least resistance, so the alternative paths fade by neglect rather than by edict. The blunt-force version of this, which gets discussed less than it should, is also valid: deprecate the legacy report, retire the manual template, close the orphan SharePoint folder. If you do not, the team will keep using them.
Why this is harder for nonprofits
Adoption is harder in mission-driven organisations than in commercial ones because the incentives sit differently. In a commercial organisation, an AI workflow that lifts revenue ten percent has an obvious champion. The person whose target it lifts. In a nonprofit, the equivalent gain often has no obvious champion, because the time saved gets absorbed back into mission work that is already under-resourced.
Executive sponsorship matters more in our sector, not less. The champions are the people whose authority makes it acceptable for staff to spend time learning a new way of working when the old way is delivering, just slower.
The gap between ninety-two and seven
Three patterns I find myself coming back to, and they all sit underneath the gap between ninety-two and seven.
Specify the champion before the technology. The internal owner needs to be agreed before the contract is signed, not after. They should be at C-suite altitude or have direct standing access to it. This part is not new.
Insist that the leadership team will use the system themselves. Not in the abstract. Specifically. Not just the CEO. The development director. The head of programmes. The board. If the honest answer is we will train the team, the project is in trouble. If the answer is I will be in the system from week one and so will my direct reports, the project has a chance.
Close the alternative paths. Decide which legacy reports, templates, and folders get retired when the new workflow goes live. The tool you do not retire is the tool the team will keep using.
The rule on champions still holds. The two new rules, that everyone in the leadership team has to be in the system themselves, and that the system has to be the path of least resistance with the alternatives closed, are what most organisations have not yet absorbed. The seven percent have. Most of the ninety-two have not. That is almost the whole story.
The technology will keep moving in three-month cycles. The human work of getting an organisation to actually adopt anything will continue to move in human time. The leaders who internalise that gap, and who are willing to be in the system themselves rather than delegating to the team, will get a lot more out of their AI investments than the ones who keep buying tools and hoping someone uses them.
Ginny Simpson is the founder of SignalWork, a strategic advisory practice that surfaces what funders, partners, and board members actually think but will not say in a meeting. SignalWork’s services include funder perception intelligence, income strategy, and AI workflow design and knowledge architecture for mission-driven organisations.
Quotes used with permission. The 2026 Nonprofit AI Adoption Report is published by Virtuous and Fundraising.AI. The Prosci data is from their Keys to Unlocking AI Adoption study, 2025.