How charities are using AI – and what trustees can learn from those getting it right

Whether you’re catching up with the news, doomscrolling on social media or meeting with friends and family, one topic keeps coming up: AI.

AI tools such as ChatGPT, Copilot and Claude are now firmly embedded in day-to-day life, and charities are following suit. The Charity Digital Skills Report 2025 found that 76% of UK charities are already using AI, up from 61% the previous year.

Every charity operates differently, supporting a wide range of beneficiaries and delivering a broad range of activities and services. As a result, there’s no ‘one-size-fits-all’ approach to using AI, which can lead either to experimentation or, in some cases, hesitation.

The focus is shifting from whether to use AI to how to use it well. A more recent report, the Future Charity Report (April 2026), draws on national surveys to assess AI governance, readiness and acceptance. Together, these reports paint a detailed picture of both opportunity and risk.

Common themes

The Digital Skills Report found that while many charities are using AI, use tends to be isolated or experimental rather than strategic – only 2% of charities are using it at a strategic level. The Future Charity Report supports this, with two thirds of charitie describing themselves as exploring or experimenting, and less than one in four having approved tools, policies or training in place.

The Digital Skills Report also found:

• Nine in 10 large charities are using AI in some form, and nearly three quarters of all charities see AI as relevant to their work
• The top three organisational uses of AI are: administration and project management (48%), preparing applications for grant funding (36%) and communications and fundraising (34%)
• The number of charities implementing an AI policy has tripled since 2024.

At the same time:

• A third of charities say that they don’t know how to get started with AI
• More than a third say their chief executive has poor AI skills, knowledge and confidence, with four in 10 saying the same of their board
• The number of charities with digital strategies in place is falling, from 50% last year to 44%, an indication that the adoption of AI is outpacing strategic planning.

Opportunities

The potential benefits of AI for charities are significant. AI can help organisations reclaim valuable time spent on resource-intensive tasks, freeing staff to focus on direct work with beneficiaries. Key opportunities include:

• Reporting and data analysis: AI tools can generate reports, analyse trends in beneficiary needs and help demonstrate impact to funders. For example, a homelessness charity might use AI to identify patterns in data that reveal unmet needs in particular areas
• Fundraising and grant applications: AI-assisted bid-writing tools can help structure applications, draft compelling cases for support and tailor proposals to funder priorities. This is particularly valuable for charities without dedicated fundraising staff, which have historically had to compete with better-resourced organisations
• Communications and marketing: AI can draft newsletters, social media content and press releases, improve email open rates and provide insights into how your content is performing
• Accessibility and inclusion: translation and transcription tools can improve accessibility and simplify complex documents
• Administration and casework: AI can summarise documents, draft correspondence, take and organise meeting notes and help manage volunteer rotas or event logistics, freeing staff and volunteer time for direct support.

Risks to manage

Alongside opportunity, charities must proceed with an appropriate level of caution. AI won’t always get things right.

When AI makes things up

Often referred to as ‘hallucinations’, AI may sometimes generate information that’s either incorrect or completely fabricated.

No one is immune from hallucinations. An international law firm recently referred itself to the Solicitors Regulation Authority after relying on a power of the court that didn’t exist. When questioned, it then provided an AI-generated response that was described as “not credible” and as having “misapplied the law”.

Technology companies haven’t been able to address this issue yet but in the interim, charities can mitigate the risk by:

• Double-checking that the source exists
• Verifying that the citation accurately reflects that source material.

When AI only tells you what you want to hear

Another risk is confirmation bias. AI can favour information that supports a particular conclusion rather than providing a balanced view. Charities can mitigate this risk by applying the following simple, structured review process:

• Assess the AI response itself: does it answer the question? Is the reasoning coherent?
• Verify the sources: do they support the conclusions? Are they current and authoritative?
• Reassess the response: in light of the sources, does the conclusion still hold?

Technically correct, practically useless

AI lacks common sense. It can follow instructions and apply logic, but it struggles with context. A widely shared example illustrates this:

A user asked AI a simple question, with an obvious answer: “I need to take my car to the car wash 100 metres from my house. Should I walk or drive?”

The response suggested walking – it’s only 100 metres, which sounds sensible until you need to drive through the car wash.

For charities, this limitation can have real consequences.

This might manifest as grant applications that tick every box but miss what a funder is really looking for, or policy documents that are legally sound but fail to reflect the charity’s values. AI can produce compliant outputs, but it can’t capture what makes a charity unique.

What trustees can learn from those getting it right

While the sector has made good progress, governance still lags behind adoption. The revised Charity Governance Code (November 2025) addresses this directly, recommending that charities have an AI and technology policy.

Trustees don’t need to understand how large language models, neural networks or machine-learning algorithms work. They need to understand what AI means for their organisation’s risk profile, data responsibilities, staff and beneficiaries.

Charities that are getting AI right share several common characteristics:

1. They have audited their current use, both formal tools adopted by the charity and informal use by staff. Understanding what’s already happening is the essential first step
2. They have bespoke AI policies. This doesn’t need to be a 50-page document. For most charities, two to four pages covering approved tools, data restrictions, review requirements, transparency, accountability and an annual review cycle is sufficient. An AI policy should reflect the charity’s values, the nature and scale of its operations and be prescriptive without being restrictive
3. Successful charities ensure human oversight of AI outputs. The risks of failing to do so have been illustrated by recent criticism of the Department for Work and Pensions, where AI tools used to detect benefit fraud showed bias according to age, disability and nationality. Any charity using AI in ways that affect beneficiaries must ensure meaningful human review is built in. No decision that materially affects a beneficiary should be made by AI alone.

The principles of good governance – duty of care, prudent management and informed decision-making – apply to AI as they do to any other strategic decision.

The questions trustees should be asking

If your board hasn’t yet engaged with AI, the following questions provide a practical starting point:

• About current use: is anyone in our organisation currently using AI tools, formally or informally? What data are they inputting into these tools? Do we have a written AI policy?
• About risk: have we conducted a data protection impact assessment for any AI tools we use? Under data protection legislation, the charity remains the data controller even when processing is done by a third-party AI tool. Could any of our AI use adversely affect beneficiaries, particularly vulnerable groups?
• About opportunity: where are our staff spending the most time on repetitive administrative tasks? Could AI help us improve our evidence base or reporting capacity?

Looking ahead

In April 2024, the Charity Commission indicated it didn’t anticipate producing specific AI guidance. No such guidance has materialised. However, the refreshed Charity Governance Code now recommends that charities have an AI policy, and the Fundraising Regulator published AI guidance in December 2025. Trustees should treat AI governance as an established compliance consideration, not a future one.

The biggest risk for most charities isn’t adopting something dangerous, but failing to engage with AI at all and falling further behind organisations that use it to serve their beneficiaries more effectively.

AI isn’t going away. The charities best placed for the years ahead will be those whose trustees asked the right questions early, put proportionate governance in place and gave their organisations room to learn.