Adoption Isn’t the Problem
What 150 conversations this spring taught me about where the nonprofit sector actually is with AI.
*I'm Ryann Miller, founder of Spark & Signal. For many years I’ve shared ideas on Linkedin — here. I'll keep doing that, but this is where I'll explore them more deeply. You'll find essays about AI, leadership and organizational change in the nonprofit and charitable sector.*
This spring I spoke at four conferences, attended another, and led four online sessions. One keynote, one full day lab, one half day workshop, and many breakout sessions. Leaving out the discussions and Q&As in the actual sessions, and just thinking the conversations before, after and in the hallways, I talked with more than 150 nonprofit leaders.
I asked almost everyone a version of the same open ended question: how does AI look and work inside your organization?
I kept it intentionally broad. I wasn’t trying to confirm a theory. I wanted to understand how people made sense of AI when I didn’t steer the conversation. What surprised me wasn’t only the answers.
It was how people interpreted the question. I’d asked about AI inside their organizations. Most people answered by talking about technology.
“We officially use Copilot, but people also use ChatGPT or Claude.”
“We have an approved tool... but I’m sure there are others.”
“I know staff are experimenting. I don’t really know with what.”
Almost nobody started with leadership, decision making, culture, organizational change or how work itself was evolving.
What struck me wasn't how many organizations were using AI. It was how little visibility leaders had into where, how and why it was being used, or what that meant for their organization.
That’s not, or not just, a technology problem. It’s a visibility problem. Technology introduces new risks, but lack of visibility is a risk in its own right. And visibility matters because it shapes what we orient ourselves around.
If leaders can’t see what’s actually happening, the conversation naturally collapses toward the visible parts inside the organization: approved tools, policies, procurement decisions and adoption metrics. It also gravitates toward the visible influences outside the organization: technology vendors and their framing, conference sessions, board priorities, and what donors or constituents are asking about. All of those things matter. And they just aren’t the whole picture.
Visibility shapes attention. Attention shapes decisions.
When leaders can only see part of what’s happening, they naturally orient around that partial picture. Once we’re orienting around the wrong things, we start measuring the wrong things.
The policy as performance
Almost everyone I spoke with was keen to tell me, later in the conversation, that their organization had an AI policy. I’d ask what happened when they encountered a question the policy didn’t answer. Almost every time, the answer was some version of, “we wing it”. I sometimes asked if they’d read it. Often they hadn’t read the final version of it..or had but didn’t know where it lived and if anyone used it.
This isn’t a criticism; it’s a signal that’s worth more attention.
A policy nobody reads, and that doesn’t help people make real decisions, isn’t real governance. Beyond some broad guardrails and checking the “our board wanted it” box, the AI policy is often performative governance. It’s a document.
Meanwhile, staff keep working. They experiment, improvise, ask colleagues, use other tools, and make judgment calls on their own. Every single person.
The gap between policy and practice is where visibility disappears. And when leaders can’t see practice, they can’t meaningfully improve it.
We’re asking the wrong questions
Most public conversations about AI in the nonprofit sector, and honestly much of the conference programming, begin in roughly the same place. Which tools? Which prompts? Who’s using what? What are the efficiency gains?
Those aren’t bad questions. But they are downstream questions. They’re wrong if they’re the first questions we ask because they sidestep purpose. The why. These types of questions reduce a complex organizational change to the parts that are easiest to see. They assume either that we’ve already answered the strategic questions, or worse — that we don’t need to. The questions I propose we start with are strategic:
What are we trying to accomplish? What should guide our decisions? Which values, people, risks and tradeoffs belong in these decisions? Where might AI genuinely help? How will we monitor, evaluate, share, debrief, learn and iterate? What must remain human? What do we need to protect and preserve as the technology changes?
Plus one of my faves: how will we know whether we’re making better decisions six months from now than we are today?
Adoption measures usage. It doesn’t measure judgment.
It doesn’t tell you whether your organization is becoming more capable of making thoughtful decisions as the environment keeps changing. This is the shift I think most organizations still need to make.
The challenge isn’t getting people to use AI. It’s building an organization that can make good decisions about AI together, consistently, as new questions continue to emerge.
What this actually requires
Organizations won’t succeed because more people use AI. They’ll succeed because more people know how to make thoughtful, consistent decisions about where AI belongs, where it doesn’t, and how those decisions evolve over time. (I’m sorry if you’ve attended one of my recent sessions and heard me say this; I’ve been a bit of a broken record on this point.)
That looks like a fundraising team with shared ways to decide and communicate which donor communications can involve AI, which require more or less human judgment, or no AI at all, when to ask for guidance, and who should make that call. Finance, HR, programs and communications will land in different places (as they should).
The goal isn't consistency in answers. It's consistency in how decisions get made. Shared ways of thinking, communicating and exercising judgment so teams can move quickly without pulling in different directions.
That’s what allows organizations to adapt without reinventing every decision from scratch. And none of that is possible without visibility.
Where I start
People often assume my work starts with AI strategy. It doesn’t. It starts by helping leaders and teams answer questions they often can’t answer confidently today.
Where is AI already changing your work? Where are decisions already being made? Who is making them? Based on what? Which decisions are working well? Which ones are inconsistent? Where are people improvising because they don’t have shared guidance?
Only once those things become visible does strategy become useful.
Not isolated decisions about tools, but shared decisions about purpose, judgment, responsibility and practice. From there, we build the structures that allow those decisions to spread across the organization instead of living inside one enthusiastic manager or one AI committee.
The work isn’t about producing a better policy. Or getting to a higher adoption rate. It's about replacing fragmented, inconsistent decision making with shared ways of thinking, deciding and adapting. That's how organizations reduce unnecessary risk, avoid duplication, build confidence across teams, and keep making better decisions as the technology changes.
For me, the work typically follows four stages:
Understand what’s actually happening, where AI is showing up, and what people are experiencing.
Decide what matters, who decides, and what good judgment looks like in your organization.
Organize those decisions into shared practices, workflows, roles and expectations.
Adapt as the technology, your organization and your understanding continue to evolve.
Skipping straight to tools, policies or rollout almost always creates the very problems organizations are trying to solve: fragmented decision making, inconsistent practice, hidden risk and teams left to improvise on their own.
The question I keep coming back to
If you discovered tomorrow that half your staff were using AI differently than you thought, would that change any leadership decisions you’re making today?
If the answer is yes, then the gap between what you think is happening and what’s actually happening deserves attention.
Not because AI is inherently dangerous. Because good decisions begin with seeing clearly.
Organizations make better decisions when they can see clearly, orient themselves around the right questions, and deliberately build the capacity to keep learning and adapting as the technology changes.
If you're interested in how nonprofits can see clearly, ask better questions and make better decisions in the age of AI, I hope you’ll subscribe.
PS I used an LLM while writing this article, intentionally and transparently. The ideas, experiences, conclusions and point of view are mine. I use LLMs to challenge my assumptions, sharpen my thinking, tease and evolve ideas (and yes, the occasional rant or vent) into something clearer and more useful, and improve how I express them.
The process looks something like: my (way too long) first draft → LLM → me → same or another LLM → me → me questioning everything → me back to focusing on the piece → me publishing the piece. (I’m not saying it’s efficient. Just honest.)