AI adoption in charities is about culture and change, not technology

There's a conversation happening across the charity sector right now about AI. Which tools should we use? Should we go with Copilot or ChatGPT? What about Claude? Do we need a chatbot? The conversation is almost entirely about technology. That's the wrong conversation.

It's worth acknowledging that almost everyone in the sector is a beginner right now. We're less than two years from ChatGPT launching, and most charities are still figuring out what AI even means for their work. That's fine. It means there's a lot of headroom to improve. But it also means we're at the stage where the organisational questions matter more than the technical ones.

The tools are not the hard part. They're getting cheaper, more capable, and more accessible every month. Free tiers are good enough for most everyday tasks. A reasonably curious person can learn the basics in an afternoon. But nobody wakes up needing "AI." They need 10 minutes back in their day, or a thinking partner to spar against, or a way to process a mountain of feedback without hiring someone. The technology is only useful when it's pulled by a real problem, not pushed by a vendor.

And this isn't new. This is how it has always worked. The reason a CRM doesn't get used is almost always cultural: it doesn't fit the needs, it doesn't fit the work. Digital transformation projects have been dressing up cultural challenges as technology problems for years. AI is no different.

We've spent the past year working with charities of different sizes on AI readiness, strategy, and training. The pattern is consistent. The technology is rarely what holds organisations back. What holds them back is the organisational stuff: how decisions get made, how risk is managed, whether teams have permission and time to experiment, and whether leadership understands what they're asking for when they say "we need an AI strategy."

Consider a scenario we've seen more than once. A senior leader reads about AI, gets excited, and asks someone to "look into it." That person does some research, maybe runs a few experiments, and comes back with ideas. Those ideas then enter a governance process designed for large technology procurements. They need a business case. A risk assessment. Board approval. Six months later, nothing has happened and the person who was asked to look into it has gone back to their actual job.

The technology worked. The governance didn't. The structures that charities use for major IT decisions are wrong for AI experimentation. AI adoption, at least in the early stages, looks more like learning a new skill than buying a new system. You wouldn't put a business case to the board before letting your fundraising team attend a training course. But many charities are treating AI experimentation with the same rigour they'd apply to a CRM migration.

Permission to experiment

The charities making progress with AI have something in common: they've given people explicit permission to try things. Not a vague "we're open to innovation" statement in a strategy document. Actual permission. A team lead saying "spend two hours this week trying AI on that task and tell me what happens." A CEO telling the board "we're running small experiments and I'll report back, but we're not waiting for a full strategy before we start."

This is harder than it sounds. Charity culture is, for good reasons, cautious. People are also skeptical, and rightly so. It's the same Silicon Valley that pushed social media and overpriced online ads. Staff who've watched the tech industry over-promise and under-deliver are not being irrational when they're dubious about the latest thing.

The most useful thing we've found is to start by asking people why they don't want to use this technology. Not to argue with them. To listen. When people feel heard about their concerns, even if they ultimately need to use the tools for their job, the resistance drops. This isn't sophisticated change management theory. It's basic respect, and it works better than enthusiasm-driven rollouts where skeptics are treated as obstacles.

Caution applied uniformly does become paralysis though. Drafting a fundraising email with ChatGPT is not the same risk category as using AI to triage safeguarding referrals. Treating them as if they require the same level of governance approval means the low-risk, high-value uses never get off the ground.

Teams stretched too thin

The other organisational reality is capacity. Most charity teams are running at or over capacity. They don't have slack in the system to experiment. And AI experimentation takes time. Not just to learn the tools, but to figure out where they fit into existing workflows, to test whether the outputs are reliable, to develop the judgment about when AI helps and when it doesn't.

When organisations tell us they want to adopt AI but nobody has time, the honest response is that this is a prioritisation decision, not a time problem. You're choosing to keep doing everything the same way. That's a legitimate choice, but it is a choice. The charities finding time for AI are usually making a deliberate decision to spend less time on something else, at least temporarily. Sometimes that means accepting a lower standard on something for a month while a team member builds a new skill. That's uncomfortable in a sector where standards matter, but it's how organisational learning actually works.

Strategy documents vs. actual change

We've seen charities produce excellent AI strategy documents that change nothing. The document gets written, presented to the board, approved, and filed. Six months later, the same people are doing the same work in the same way.

A strategy document is not change. It's a description of intended change. The gap between those two things is where most AI adoption efforts stall. Closing that gap requires someone with enough authority and time to keep pushing. It requires regular check-ins on whether anything is actually different. It requires accepting that the strategy will need to change as the organisation learns what works.

The best approach we've seen is to start experimenting before you write the strategy. Run a few small pilots. Let three or four teams try AI on real tasks for a month. Gather what they learn. Then write a strategy informed by actual experience rather than assumptions about what AI might do. The strategy will be better, and you'll already have people inside the organisation who've used the tools and can champion them.

Any change needs the push away from the current way of doing things, combined with the pull of something better, to be stronger than the comfort of habit and the anxiety of the unknown. That's the culture work: building a bridge between how things are done now and how they could be done. It means figuring out the actual problems people have and applying solutions to those problems. Not "we need AI" but "we need to stop spending two days a month on this task."

The charities that get somewhere will be the ones asking "what problem are we actually solving?" rather than "which tool should we buy?"