The 90% Problem: UK Charities Are Using AI in Their CRMs But Barely Moving the Needle on Retention

The headline number is striking. According to Blackbaud's ninth annual Status of UK Fundraising report, published this spring, 91% of charities now use AI in some form. Read that again: nine in ten. And yet the same research makes clear that advanced applications - personalisation at scale, predictive ask modelling, retention analytics - remain rare. Only around 7% of organisations describe a major improvement in what AI has actually enabled them to accomplish.

That gap is where development teams should be spending their energy this summer.

What the data actually shows

Blackbaud's own institute has been tracking the divide for months. Its June 2026 'Bridging the AI Effectiveness Gap' research identifies a cohort it calls 'AI-Adaptive' organisations: the 10% that have moved beyond scattered, individual use towards systemic, governed deployment across fundraising operations. That top 10% are seeing better donor retention and higher fundraising revenue, rising in step with their AI maturity. The organisations below them, the vast majority, are using AI to do things faster, but not differently enough to change outcomes.

The mechanism matters. An AI layer sitting on a clean CRM can log new gifts, deduplicate contacts, apply the right consent and Gift Aid flags, segment supporters by lifetime value, recency and channel, and trigger personalised thank-you communications within minutes of a gift landing. That is not futurism; Struan.ai is already marketing exactly this proposition to small UK charities, billing it as operational depth that used to require a dedicated fundraising and IT team. The real constraint is not the tool. It is the data underneath it.

Quematics, a UK data consultancy working across the voluntary and community sector, put it plainly in a piece published last month: 88% of UK charities report using AI tools in their day-to-day work, but only around 7% see major impact from it, and just 46% describe their use as active or strategic. The reason is almost always the same. Messy CRM data leads to inaccurate forecasting, missed opportunities, and compliance exposure. Garbage in, personalised garbage out.

What the platform market actually offers

For UK charities evaluating what is available, the market is more fragmented than vendor pitches suggest. Plinth, the only UK-built CRM to offer AI natively across case management, grant assessment, due diligence, impact reporting, and donor stewardship, hosts data in the UK and EU and is used by over 60 funders managing more than £200 million in annual grants. Blackbaud's Raiser's Edge NXT has introduced donor analytics and fundraising optimisation features, though its data hosting defaults to the US, a live GDPR consideration for any UK data protection officer. Beacon and Donorfy, two of the most popular UK-native platforms for individual giving, have yet to introduce substantive AI capabilities.

Salesforce Einstein offers enterprise-grade AI but requires significant investment and configuration that most mid-sized charities cannot justify. For most development teams, AI stewardship today means layering a third-party tool or workflow on top of an existing CRM, not buying a pre-integrated solution.

The stewardship stakes are rising

None of this is academic. CAF's UK Giving Report 2025 found the median monthly donation has risen from £20 in 2019 to £28, while the mean has climbed from £46 to £72. Higher individual gifts mean stewardship matters more, not less. Exceptional gifts, large one-off donations, have driven income growth for four years running, precisely because they reflect long-term investment in tailored supporter experience. Moving from basic RFM (recency, frequency, monetary) segmentation to behavioural data that captures petition signatures, event attendance and volunteer hours is where AI-assisted CRM earns its cost.

The trademark filing picture adds an interesting footnote. AIBD analysis of IPO (TMD) data, as of August 2026, shows just 835 Class 36 trademark filings in Q3 2026, down 65% on the prior period. Class 36 covers financial services and fundraising activities. That is a sharp contraction in brand-IP activity in the very space where AI-powered donor products are proliferating: a signal that vendors are either consolidating, or that genuinely novel propositions are thinner on the ground than the marketing suggests.

The trust question the Regulator has already answered

Development directors who assumed they could quietly deploy generative AI for appeal copy and stewardship communications should have read the Fundraising Regulator's AI guidance, published in December 2025 and increasingly cited in compliance reviews. The regulator's position is unambiguous: charities remain accountable for all AI use in fundraising, including outputs generated by third parties. Trustees must exercise active oversight. Proportionate risk assessments should precede deployment. Charities should publish their AI fundraising policies on their websites to maintain trust and transparency with donors.

The regulator was equally clear on labelling. Not every AI-assisted communication needs a disclosure tag, but the guidance is explicit that where AI-generated content could reasonably mislead donors, through realistic imagery, synthetic voices, or chatbot solicitations, the greater the risk, the more transparent the charity must be. That standard applies directly to personalised stewardship emails, AI-drafted legacy conversations, and any chatbot fielding donor queries.

YouGov's 2026 consumer research, reviewed in trade press this month, found that 69% of people are worried that AI-generated content may be misleading, with 67% concerned about their inability to distinguish machine-created from human content. For charities, whose entire income proposition rests on emotional trust, that is a retention lever pointing the wrong way if handled carelessly.

Charities that have tried AI-generated imagery and then stopped, citing ethical, reputational or authenticity concerns, have already learned this lesson at cost.

What your development team should do this week

First, audit the data before touching the AI. If your CRM has duplicate records, missing consent logs, or inconsistent Gift Aid flags, no personalisation layer will save you, and the ICO will not accept 'the AI did it' as a mitigation.

Second, get a board paper on AI governance to your next trustee meeting. The Fundraising Regulator has made trustee accountability explicit. If your board does not know what AI is being used for stewardship communications, that is a governance gap a complaint could expose quickly.

Third, draft and publish an AI fundraising policy before your autumn appeals go out. It does not need to be long. It needs to exist, to be honest, and to be on your website. The charities that will struggle are the ones that treat this as admin rather than donor relationship management.

Fourth, review how your CRM handles behavioural segmentation. Move beyond gift history. Volunteer hours, campaign responses, event attendance: these are the signals that separate a personalised stewardship journey from a mail-merge with a first name.

The AI maturity dividend is real. But it compounds only for organisations that have done the unglamorous work first.