6. The AI Skills Gap in UK Charities Is Not What You Think

The charity sector has a skills gap problem. Everyone agrees on this. The Charity Digital Skills Report 2025 found that 50% of charities cite lack of skills and expertise as a barrier to digital progress. Thirty-five percent rate themselves as poor AI users. Twenty-nine percent don't use AI at all. And 65% say that more training for leaders and trustees would help.

The sector's response has been predictable: more training courses. Microsoft and LinkedIn have launched free learning paths. TechSoup offers on-demand AI courses. Charity Digital runs workshops. CAST has published step-by-step guides. Charity Excellence has built an entire free training programme funded by Microsoft. NCVO ran an AI essentials session during Trustees' Week. NetHope's Unlocking AI for Nonprofits series has over 5,000 enrolments.

And yet, after all of this training, the 2025 report also found that 64% of charities still make limited or no use of AI in their day-to-day work. The 2026 data shows improvement — 88% now use AI — but still only 46% are doing so actively or strategically.

Something doesn't add up. If the problem is skills, and the sector has invested heavily in training, why isn't it working faster?

We want to suggest something that might be uncomfortable: the skills gap, as commonly understood, is largely a misdiagnosis. The barrier facing most charity workers isn't that they can't use AI. It's that they don't feel they're allowed to, they don't know where to start in a way that's relevant to their actual job, and nobody's given them a framework that turns occasional experimentation into consistent practice.

The real gap isn't skills. It's confidence, permission, and structure.

The confidence problem

Here's a test. Take any charity worker who owns a smartphone and ask them to open ChatGPT (free version, no account needed on the app) and type: "Summarise this text into three bullet points" — then paste a paragraph from their last board report.

They'll get a usable result in about ten seconds. No training course required.

The basic capability to use generative AI tools is trivially easy to acquire. The tools are designed to be used through natural language. You type what you want in plain English. The barrier isn't "I don't know how to use this tool." It's "I'm not sure I should be using this tool, I'm not confident the output will be good enough, and I don't want to look like I don't know what I'm doing if I get it wrong."

That's a confidence problem, not a skills problem. And confidence problems aren't solved by more training courses. They're solved by safe environments to practise, clear permission from leadership, and quick wins that demonstrate value.

The CDSR data supports this reading. When charities say they have "poor AI skills," they're not saying they've tried and failed. They're saying they haven't really tried. The self-assessment is measuring perceived capability, not actual capability. And perception is shaped by anxiety, not experience.

The permission problem

Forty-three percent of nonprofits rely on just one or two staff members for all IT and AI decision-making, according to the TechSoup 2025 report. In most small charities, that person is the CEO — who is also managing fundraising, governance, HR, service delivery, and everything else.

When nobody in the organisation has explicitly said "it's okay to use AI for your work," most staff won't. Not because they're uninterested — the data shows strong appetite — but because charity culture is inherently cautious. Staff worry about data protection (as we covered in Episode 4, usually unnecessarily for Tier 1 use). They worry about appearing lazy — that using AI means they're not doing their job properly. They worry about quality — that AI outputs will embarrass the organisation.

This is a permission problem. The solution isn't a training course. It's a one-page AI policy (Episode 1 in this series) that explicitly says: "We support the use of AI tools for the following tasks. Here's what's approved. Here's what's not. Go ahead."

An extraordinary amount of latent capability unlocks the moment an organisation gives its staff formal permission to experiment. No training required. Just a sentence from leadership that says "yes."

The structure problem

The third gap is the hardest to close, and it's the one that actually matters for long-term adoption.

Most charities that use AI are at what the CDSR calls the "informal" or "exploratory" stage. Someone on the team uses ChatGPT occasionally. They've found it helpful for a few tasks. But there's no shared approach, no consistent practice, and no way to build on what individuals have learned.

This is the gap between "I use AI sometimes" and "we use AI as an organisation." Closing it requires structure — not training in the traditional sense, but a framework that answers practical questions. Which tasks should we prioritise for AI? What prompts work well for our specific workflows? How do we share what works across the team? How do we build AI into our processes so it's not dependent on one person's enthusiasm?

This is genuinely where most organisations get stuck, because the jump from individual experimentation to organisational adoption requires someone to do the work of translating generic AI capability into sector-specific, role-specific practice. And that someone is usually the same overloaded CEO who's already managing everything else.

What actually works instead of more training

If the problem is confidence, permission, and structure rather than raw skills, the interventions look different.

For the confidence gap: start with one task, this week. Don't train people in the abstract. Pick the single task that consumes the most admin time in your team — it's usually meeting notes, first-draft communications, or report formatting — and ask one person to use AI for it. One task, one person, one week. Debrief on what happened. Did it save time? Was the output useful? What needed editing? This creates evidence that's more persuasive than any training course, because it's evidence from your own organisation, about your own work.

For the permission gap: publish your AI policy. Episode 1 of this series includes a one-paragraph version you can implement today. The act of publishing a policy — even a short one — transforms AI from "something people do secretly" into "something the organisation endorses." Staff who were hesitant will start experimenting within days. The CDSR data backs this up: charities at the "advancing" and "advanced" stages of digital are far more likely to have policies in place. The policy enables the adoption, not the other way round.

For the structure gap: use a prompt library. The single most practical intervention for moving from "I sometimes use ChatGPT" to "we consistently use AI well" is a shared collection of prompts tailored to your actual workflows. Not generic prompts from a blog post — prompts that reference your specific tasks, your reporting templates, your tone of voice.

When a new staff member can open a shared document, find a prompt labelled "First draft: quarterly funder report," paste in their data, and get a structured output in three minutes — that's organisational capability. It doesn't depend on individual enthusiasm. It survives staff turnover. And it compounds over time as the team adds prompts for new tasks.

For all three gaps: designate an AI lead. Not a full-time role — just someone who's willing to be the first point of contact for questions, to collect and share what's working, and to keep the policy and prompt library up to date. In most small charities, this is the person who's already experimenting with AI informally. Naming them gives the work legitimacy and creates accountability for progress.

Where training does help

We’re not arguing that training is worthless. But it helps in specific, targeted ways — not as the default response to every AI challenge.

Training is useful for trustees and senior leaders who need enough understanding to govern AI effectively (see Episode 3). The free resources here are strong — NCVO's AI Essentials for Trustees, Microsoft's LinkedIn learning paths, CAST's guides, and Charity Excellence's training programme are all worth recommending.

Training is useful for staff who are moving beyond basic use into more complex tasks like data analysis or workflow automation, where the tool's capabilities are less intuitive.

And training is useful for building prompt engineering skills — the ability to write instructions that get consistently good outputs from AI. This is genuinely a skill that improves with practice and guidance, and it's the one area where the "skills gap" framing is accurate.

But for the 64% of charities making limited or no use of AI, the bottleneck isn't training. It's the absence of permission, confidence, and a practical framework for getting started. Fix those three things and the "skills gap" shrinks dramatically — because the skills were never really the problem.

Free resources worth knowing about

If you do want to build AI knowledge across your team, these are the strongest free resources available to UK charities right now:

CAST's AI Resources and Shared Digital Guides — step-by-step practical guides written by charities for charities, covering everyday AI use cases. wearecast.org.uk

Charity Excellence's Free AI Training Programme — funded by Microsoft, requires no prior digital expertise, covers everything from awareness to governance to practical use. charityexcellence.co.uk

TechSoup's AI courses and webinars — on-demand practical training including Microsoft Copilot for charities. techsoup.org

Microsoft's AI Skills on LinkedIn Learning — free, self-paced courses covering core AI concepts. Endorsed by NCVO.

Charity Digital's AI Hub — articles, webinars, and podcast episodes covering AI adoption in the sector. charitydigital.org.uk

NCVO's AI and small charities resources — guidance and event recordings focused on small charity needs. ncvo.org.uk

The bottom line

The sector's investment in AI training is welcome. But training alone won't close the gap if the real barriers are confidence, permission, and structure.

Give staff permission to experiment. Publish a policy. Start with one task. Build a shared prompt library. Name an AI lead. These five things will do more for your charity's AI adoption than any training course — and none of them cost anything.

Our free AI Readiness Audit helps you identify which of these barriers is strongest in your organisation. And the toolkit provides the structure layer — a readiness assessment, a use case discovery process, a prompt library with 50+ charity-specific prompts, and a 12-week implementation roadmap that turns experimentation into practice. But start with the permission. Everything follows from that.