9. What AI Adoption Actually Looks Like Inside a Charity

Here's a statistic that made headlines across the charity sector this year: Around 80% of UK charities now use AI tools in their day-to-day work. Up from 76% the year before and 61% the year before that.

And here's the statistic that should have made headlines but didn't: only 3% of charity trustees said their charity was using AI at all.

That gap — between organisations reporting AI use and trustees knowing about it — tells you everything you need to know about what "adoption" actually looks like in most charities. It's not a strategy. It's not a transformation programme. It's one person using ChatGPT at their desk, without telling anyone.

What "using AI" actually means

The 2025 Charity Digital Skills Report broke adoption into stages for the first time. Of the 76% using AI that year, only 2% were using it strategically or planning to scale. Fifty-one percent were "exploring" — which is the survey's polite term for informal, individual use. Twenty-three percent were piloting.

The 2026 early data shows improvement: 46% are now "actively or strategically" using AI, nearly double the previous year. But that still means more than half of charities reporting AI use are at the exploratory stage — one person, one tool, no shared approach.

The Charity Excellence AI Survey 2026 put it more bluntly: "AI is being used informally, without anyone really talking about it." Most use flows through individual staff or volunteers working with tools like ChatGPT or Microsoft Copilot on their own initiative. It's happening without policies, without board awareness, and without any measurement of whether it's actually helping.

This matters because the headline statistic creates a false sense of progress. When a CEO reads "80% of charities use AI," they might conclude their organisation is behind if it hasn't implemented an AI strategy. The reality is that most of these organisations haven't implemented a strategy either. They have a staff member who's quietly using ChatGPT to write newsletters.

The five stages of charity AI adoption

Based on the data and what we see in practice, charity AI use falls into five recognisable stages. Most organisations are at Stage 1 or 2.

Stage 1: Invisible use. Individual staff members use AI tools on personal devices without organisational knowledge. No policy. No board awareness. The Charity Commission's Trust in Charities research found just 3% of trustees knew their charity used AI — suggesting invisible use is extraordinarily common. The risk here isn't that AI is being misused (most informal use is low-risk Tier 1 activity). The risk is that the organisation can't learn from it, can't build on it, and can't manage it if something does go wrong.

Stage 2: Acknowledged but unstructured. The organisation knows AI is being used. There might be a policy, or at least an informal understanding. But there's no shared approach — no prompt library, no designated lead, no measurement of impact. Individual experiments happen but knowledge stays locked in individual heads. Most of the charities in the "exploring" category sit here. The jump from Stage 1 to Stage 2 is small — it requires only that someone says "we know this is happening and we support it." But the jump from Stage 2 to Stage 3 is harder.

Stage 3: Structured experimentation. The organisation has a policy, a shared prompt library, and at least one person responsible for coordinating AI use. Specific tasks have been identified and trialled. There's some measurement of time saved or quality improved. The team shares what works. This is what the CDSR calls "piloting and using" — and it's where the real value starts to compound, because knowledge becomes organisational rather than individual.

Stage 4: Embedded practice. AI is part of how the organisation works, not an experiment running alongside normal operations. Processes have been redesigned to include AI — grant reports are structured using AI as standard, communications go through an AI drafting step, meeting notes are AI-assisted as routine. Staff expect to use AI. New starters are inducted into the prompt library. The organisation can quantify the time saved and redirect that capacity to mission.

Stage 5: Strategic integration. AI informs organisational strategy, not just operations. Service delivery is enhanced by AI tools. Data analysis drives decision-making. AI is part of how the charity plans, not just how it executes. Only 4% of charities reported strategic AI use in 2025. The 2026 data suggests this is growing, but for most small charities, Stage 5 is a horizon rather than a near-term target.

Where are you? And does it matter?

Most charities reading this will recognise themselves in Stage 1 or Stage 2. That's fine. The question isn't "how do we get to Stage 5?" — it's "what does it take to get from where we are to the next stage?"

From Stage 1 to Stage 2: publish a policy. Tell staff it's okay. Acknowledge what's happening. (Episode 1 in this series.)

From Stage 2 to Stage 3: build a shared prompt library. Designate an AI lead. Try one task formally with measurement. (Episode 6.)

From Stage 3 to Stage 4: redesign specific workflows to include AI as standard. Build AI into induction. Measure and report to the board. (Episode 7.)

Each jump is smaller than it looks. And critically, you don't need to reach Stage 5 to capture most of the value. The biggest gains — recovered admin time, faster drafting, better-structured reports — come between Stage 2 and Stage 3. That's where effort is best concentrated for a small charity.

The governance gap is the real story

If AI use is near-universal in the sector but trustee awareness is in single digits, the governance gap is enormous. The Charity Governance Code was refreshed in November 2025, explicitly addressing technology and AI. But most boards haven't had the conversation. Most organisations don't have policies. Most informal use is invisible to the people responsible for governance.

This creates three specific risks. First: data protection. If staff are entering data into AI tools without organisational oversight, nobody's checking whether personal data is involved. Most informal use is low-risk — but nobody knows, because nobody's asked. Second: quality. If AI-assisted outputs are going to funders, beneficiaries, or the public without review processes, errors are a matter of time. Third: missed opportunity. If twenty different staff members across an organisation are each using AI individually but never sharing what works, the organisation is capturing a fraction of the potential value.

The gap between use and oversight isn't a technology problem. It's a leadership problem. And it closes with three steps: a policy (permission and boundaries), a prompt library (shared knowledge), and a board conversation (governance). Those three steps move an organisation from Stage 1 to Stage 3 — which is where the value lives.

The sector in twelve months

Looking at the trajectory — it's reasonable to expect near-universal AI use by the end of 2026. The question for the sector isn't whether charities will use AI. It's whether they'll use it well.

The 2026 early data offers reasons for optimism. The adoption gap between small and large charities has effectively closed.

But the governance gap remains wide. Closing it doesn't require technology expertise. It requires organisational leadership — someone willing to acknowledge what's happening, put a framework around it, and build from there.

If this series has given you that framework, you have everything you need. Our free AI Readiness Audit tells you which stage you're at right now. And the toolkit provides the structured path from wherever you are to the next stage — readiness assessment, use case discovery, prompt library, governance pack, ROI calculator, and a 12-week implementation roadmap. But the first step is always the same: acknowledge what's already happening, and decide to do it well.