Report30th September 2026
See the research questions
- Do charities have AI policies, and what approach do they take?
- Are they providing training and guidance?
- Do content practitioners feel supported in using AI?
Over half do not have a formal AI policy
- Only 38% of respondents said their organisation has a formal written policy on AI.
- A majority (56%) have either no policy or one that’s still in progress:
- 13% have informal guidance only
- 18% said a policy is in development
- 18% said there’s no policy but conversations are happening
- 7% said it hasn’t come up at all
Formal training is not widespread
87%
have not had any formal training on AI
- Only 13% of respondents said that their organisation has provided formal training on how to use AI tools.
- 42% said that there has been some informal guidance or resources.
Half do not feel supported when it comes to AI
- 49% feel somewhat or very unsupported by their organisation in navigating AI in content work.
- Only 5% said that they feel very supported by their organisation.
Deep dive: Peer communities of practice are an excellent source of support and learning
A small number of participants mentioned that they are part of communities of practice groups focused on AI. These vary from small, informal peer-led groups, to formal organisation-wide cohorts. Where they exist, communities of practice were almost universally-praised. People credited them with:
- surfacing what’s actually working when it comes to AI
- creating a space where fragmented individual practice can evolve into shared team ways of working
- reducing anxiety, shame and isolation around AI use
For example, one participant shared that they are part of a peer-led community of practice in their organisation’s marketing and communications department. The members of the community meet regularly to share learnings, prompts and use cases. They have found the group to be a useful forum for hearing different perspectives on AI, as well as comparing their own learnings.
Another participant who is part of two different AI communities of practice in their organisation, said that they value these spaces because they allow people to discuss the benefits, risks, and concerns regarding AI, helping to avoid ‘spiraling about doom and gloom’ or adopting an overly optimistic, ‘tech bro’ mindset.
CAST’s AI survey 2026 also found that ‘86% of respondents wanted access to peers within the sector to discuss AI’ and ‘62% of those who responded… wanted to see coordination between charities’.
Analysis: Governance often feels tick-box or disconnected from practice
AI governance maturity seems to vary widely across the sector. In the study, respondents reported everything from having no formal policy at all, to generic guidance that’s rarely referred to in practice, to bespoke, consultative policies developed with staff input. There were even instances where, within a single organisation, the team-level policy was more developed than the organisation-wide one it sits under.
The study echoes the findings of the Charity Digital Skills Report:
- 44% of charities are not taking action to manage risks and progress with AI
- 51% of charities do not have an AI policy
- 19% are reviewing data protection, security and safeguarding in order to progress with AI and manage risks
- 34% are making strengthening digital, data and AI governance and decision making a priority
One sub-theme was that, for some participants, their organisation’s AI policy and governance feels very disconnected from the day to day ways AI is being used, and more like a box checking exercise than something that’s usable.
The policy we have was written without involvement from our digital teams… it feels more like a tick-box exercise.
The study also found a few instances where people are using personally paid-for or otherwise unofficial AI tools and workflows because the official options haven’t caught up with what they want to do. 18% of respondents say that they personally pay for AI tools they use at work (15% regularly, 3% occasionally), and a further 7% have done so in the past. This is notable given that 85% say their organisation provides access to at least one official AI tool (53% one tool, 32% a few). This ‘shadow’ use isn’t because there’s no tool in place, it’s more about tool preference or capability gaps.
I’ll be honest, I cheat and I… pay for Claude… I can’t believe I’m the only person who’s using their own version.
When there’s a gap for governance, training and support on AI, individual actions and judgement fill that gap. People bring in their own tools, come up with their own rules, and set their own ethical boundaries. This isn’t always a bad thing – there’s plenty of strong individual practice around AI – but it comes with risks. The risks falls into two main categories:
- Exposure for the organisation: there could be breaches of data and privacy, increased risk of sharing misinformation publicly, embarrassing values and mission clashes.
- Burden on individuals: having to absorb tool costs (or not being able to afford tools that others can), concerns over making decisions that are too high level, cognitive surrender and deskilling.
Thoughtful, considered, practical governance is an enabler. It doesn’t just help to mitigate the organisation against the risks, it gives individuals the foundations they need to build good practice on.
It’s important to remember that we’re still in the early days of generative AI. Governing AI well requires organisational capacity: time to evaluate tools, legal and IT input, and confidence about a technology that’s moving fast and being marketed aggressively. Leaders are managing this alongside other competing pressures and priorities, and without some of the unique first-hand exposure that content practitioners have.
The gap between content practitioners’ concerns and leadership engagement isn’t a sign that leaders don’t care. It reflects that this is hard to get right, and that most organisations are still working out what good governance looks like.
Takeaways: How to fill a governance gap around AI
- Use your influence. If your organisation doesn’t have an AI policy in place, use your influence and speak to the person who’s responsible for this. Explain the risks of not having a robust policy and the opportunities you can see in developing one.
- Develop a policy or guidelines just for content. You could create a policy or guidelines just for content. Even if your organisation has a policy in place, you might find that you need something that deals with content-specific scenarios, use cases and risks. William Joseph’s James Gadsby Peet has a very helpful presentation on developing an AI policy.
- Start or join a community of practice: Start or join a community of practice for AI in your organisation, department or team. These groups can be a great informal governance mechanism, as well as a source of valuable, bolstering peer support. You could also look beyond your organisation for groups in the wider sector.
Starting an AI community of practice
A community of practice is a group of people connected by a shared interest in something they actually do – and who get better at it by learning from each other over time.
Three things make a community of practice distinct from a working group, a mailing list, or a team meeting:
- A shared domain. Something everyone in the group has in common, whether that’s a skill, a responsibility, or an area of work. In this case: AI.
- Genuine community. Members who connect, share, and build relationships over time. A channel in Teams or a list of people with similar job titles is not enough on its own. The interaction is what creates the community.
- Practice. Members are doing the thing, not just interested in it. Over time, they build up a shared repertoire: resources, ways of working, stories, tools, things that have and haven’t worked. That collective knowledge is what makes the community of practice useful.
It’s also worth being clear about what a community of practice is not. It is not a project working group. It doesn’t have deliverables or targets in the conventional sense. It’s not a consultation exercise, and it shouldn’t replace existing team structures.
Questions to help you develop your community of practice
- What is the domain? What specifically is the shared practice you’re bringing people together around? ‘AI’ is broad – is it all AI usage, or are you starting with a narrower focus, like AI and content, AI and comms, or AI in a specific directorate or project? Being clearer about the domain early helps people understand whether the community is for them.
- Who is the community for? Who across the organisation is using AI? Which teams, roles, or individuals are most likely to benefit from a space like this? Is there anyone who is already doing interesting things with AI that you could build around? Who do you most need to reach, and who might champion it within their own teams?
- What problem does it solve for members? Why would someone choose to come? What would they get out of it that they can’t get elsewhere? The community needs to be genuinely useful to members. What’s the honest answer to ‘why should I make time for this?’
- What does participation look like? How often would the community meet? What format – a call, an in-person session, something async? What happens between meetings? Is there a channel, a shared space? How formal or informal should the tone be? Would all members attend the same thing, or are there different tiers of involvement? How do you track and share learnings?
- Who facilitates, and how? Who will lead? What does facilitation involve – preparing sessions, keeping a record, following up, growing membership? How do you avoid this becoming a task that’s easy to deprioritise when things get busy? And how does facilitation gradually shift towards members over time?
- What does success look like, and how will you know? It’s worth agreeing early on how you’ll track whether the community is working – not through vanity metrics like headcount, but through signs of genuine value. Are members showing up consistently? Are they contributing, not just attending? Are they telling you it’s useful? Are things from the community finding their way into people’s practice? You don’t need a formal measurement framework at the start, but having a rough sense of what ‘working’ looks like will help you make good decisions about the community as it develops.
Read the report section-by-section
Appendix, reading list and resources
Introduction
Part 1: AI, organic search traffic and content discovery
Part 2: How charity content practitioners use AI
Part 3: The ethics of AI for charities
About the study
This report draws on primary research carried out in summer 2026. It combines a sector-wide survey of 63 people who produce content, mostly in UK charities and non-profits; 8 in-depth interviews with content leads; and 3 diary studies, in which practitioners logged their use of AI over a week. Survey responses were analysed quantitatively, and free-text answers, interview transcripts and diary entries were coded into themes to find patterns across all three methods.
It was produced with support from William Joseph.