Episode 2 — “The AI Adoption Paradox, Unpacked“

92% of nonprofits use AI. Only 7% say it’s moved the needle. This episode unpacks why — with three real case studies from my own client work: a donor segmentation model that changed a crisis campaign’s outreach, a development health diagnostic that turns vague problems into a ranked action list, and a campaign restructure that compressed a timeline by months.

The real difference between the 92% and the 7%? Individual experimentation vs. real organizational infrastructure. I break down what that looks like in practice, plus a simple test to check where your own team stands.

Source: The 2026 Nonprofit AI Adoption Report, Virtuous & Fundraising.AI (Feb 2026).

Sean is the founder of Magnolia Philanthropic Services, advising nonprofits nationwide on development strategy, with 25+ years in the field.

Transcript

SPEAKER_00 0:04

Last week I opened the show with a number that I never actually explained. 92% of nonprofits are using AI right now. Only 7% say it's made a real difference to what they can actually do. I got a few messages after that episode, and they all asked some version of the same thing. Okay, so what's actually happening in that gap? That's the whole episode today. Not the stat, the mechanism behind it. Why adoption is everywhere and impact is rare, and what the organizations in that lucky 7% are actually doing differently. I've got three real examples for my own client work to walk through. So this is not theoretical. This is navigating the base. And let's get into it. So let's get specific about where this number comes from because I want to be careful here. I'm not going to just repeat a stat without telling you where it comes from. So this stat comes from the 2026 Nonprofit AI Adoption Report. It was published jointly by Virtuous and Fundraising AI. And it's based on a survey of 346 nonprofits conducted this past December. And the 92 and 7 headline is actually the least interesting part of it. Here's what's underneath it. So it's not that AI is doing nothing. Most organizations are getting something out of it. But small to moderate efficiency gains and organizational transformation are two very different things. And only 7% are seeing the second one. Here's the number that actually explains the gap, in my opinion. 81% of nonprofits using AI are doing it individually. That means that one staff member has Chat GPT open in a tab, or one development associate figured out a prompt that helps them draft acknowledgement letters faster, with no shared workflow, no institutional process, and nothing that survives that person leaving. And 47% of nonprofits have no AI governance policy at all. No guidance on what data can go into a tool, no standard for how outputs get checked, nothing. Put those two numbers together and you get the paradox. Adoption is high because the barrier to an individual trying a free tool is basically zero. Impact is low because almost none of that individual experimentation is being turned into organizational infrastructure. It's a thousand people quietly reinventing the same wheel in isolation instead of an organization building one wheel together. I want to make this real. So let me walk through three situations for my own client work where this has played out, where the difference between individual experimentation and real infrastructure was the whole story. Case study one donor segmentation at scale. So I worked with a health-focused nonprofit responding to a natural disaster, trying to figure out how to talk to their existing donor base about the new urgent need. The list was over 1,100 unique donors. The old way to handle this, the individual AI use way, would have been someone on staff using AI to draft one nice email and blasting it out to everyone. You may have already done this yourself. Instead, we built a five-tier affinity segmentation model. AI helped us process giving history, engagement patterns, and past response behavior across the entire list and sort it into donors within tiers based on likely affinity and capacity for this specific asset. A major donor with a history of responding to crises appeals got a different, more personal approach than someone who'd given once five years ago to a completely unrelated campaign. That's the difference in one sentence. Individual AI use drafts want email faster. Infrastructure level AI use tells you which donors need which message in the first place. The first is a productivity tweak. The second changes the outcome of the campaign. A lot of small and mid-sized nonprofits know something is wrong with their fundraising operation, but they can't articulate what it is specifically. That vague sense of we should be doing better than this doesn't give anyone something actionable to work on. So I built a scoring tool, and I call it a development health score, that evaluates an organization across 12 distinct dimensions of development capacity. Things like donor pipeline health, gift officer capacity, data hygiene, stewardship consistency. AI does the heavy lifting of processing an organization's actual data and benchmarking it. And what comes out on the other side isn't a vague impression, it's a specific ranked list. Here are your three biggest gaps in order, and here's roughly what closing each one is worth. That's infrastructure. It's repeatable. It doesn't depend on one person's intuition, and a board can actually act on it. Here's case study three: compressing a campaign timeline. I had a client running a fundraising campaign originally planned to run across roughly nine to ten months. Business needs changed, and the real deadline moved up by several months. The individual AI use response here would be have AI help write things faster. The infrastructure response was different. We used AI to actually restructure the campaign's phases, figuring out which stages of donor cultivation, ask, and stewardship could run in parallel instead of sequentially without damaging the donor relationships that those stages are supposed to protect. The campaign timeline compressed by roughly three months, not because anyone worked three months harder, but because of the structure of the work changed. Three different situations, three different problems, but the same underlying pattern in all of them. AI didn't replace judgment in any of these cases. It gave the judgment something better to work with. More information, organized more usefully, faster than a person could do it manually. That's what separates the 7% from the 92%. So if you're listening to this and you're a development director or an ED, here's the honest question to ask your own organizations this week. Is your AI use individual or is it infrastructure? A quick blood test. If the one person on your team who's good with AI tools left tomorrow, would anything about how your organization operates change? If the answer is yes, if a process would break, a report wouldn't get made, a workflow would just stop, that's a sign you're still in the 81%, not the 7%. Moving from individual to infrastructure doesn't require a massive budget or a 12-month transformation project. It usually starts with one honest, specific question. What's the one thing your team does manually, repeatedly, that eats real time? And could that become a standard, documented, AI-assisted workflow instead of living in one person's head? That's genuinely the whole exercise. Not adopt more AI. Pick one process, make it real infrastructure instead of one person's personal trick. That's the paradox, unpacked. Adoption is easy, infrastructure is hard, and the 7% aren't smarter or better funded. They've just turned individual experimentation into something the whole organization can rely on. Next week, I want to go deeper into one of these specifically building a donor segmentation model from scratch, step by step, so you can actually see how it's built, not just hear about the outcome. I'm Sean. This has been Navigating the Maze. If this episode was useful, send it to one development director who needs to hear it. See you next week.