Seven in Ten Homecare Providers Now Use AI - But Only Half Have Any Written Policy to Govern It

Adoption Without a Safety Net

The headline number is striking enough. Seven in ten UK homecare providers are now using artificial intelligence in their operations, with that figure set to climb to 85% within twelve months. But the report buried beneath that statistic is the one that should concentrate minds in Whitehall and at the CQC: only 43% of those providers have a written policy governing how that AI is actually used.

The findings come from Birdie, a homecare technology company, whose report Moving Faster than the Rules: AI, Care Quality and the Homecare Sector in 2026 surveyed 122 homecare providers across the UK. The title is not hyperbole. It is a diagnosis.

Care Plans Written by Chatbot

What makes the governance gap particularly acute is what the AI is being used for. This is not simply a story about back-office admin tools drafting shift rosters. Around half of providers are using AI to help shape care itself, including care plans and risk assessments: decisions that carry direct clinical weight for often vulnerable clients. And the tools doing that work are largely not purpose-built for healthcare. ChatGPT is used by 63% of AI-adopting providers, Microsoft Copilot by 47%, and Google Gemini by 38%. General-purpose consumer tools, originally designed for drafting emails and summarising meeting notes, are now influencing how a carer prioritises a morning visit to a 78-year-old living alone.

That said, the outcomes data is genuinely encouraging. Three-quarters of providers (76%) say AI has improved the quality of care they deliver. Among those re-inspected by the CQC since adopting AI, 59% saw their rating improve. None reported a decline. Those are not trivial numbers. The CQC inspection framework is not designed to be gamed; an improvement in rating typically reflects real changes in how care is delivered and documented.

So the technology is working. The question is whether the legal and governance architecture around it is fit for purpose.

A Two-Tier Market in the Making

The Birdie data also surfaces something that should concern commissioners and policymakers: a growing gap between large and small providers. AI adoption sits at around 15% among the smallest homecare agencies, rising to roughly 80% among those supporting 141 or more clients. Smaller agencies told Birdie they are holding back not because they are uninterested, but because they are uncertain which tools to trust. That is a reasonable thing to be uncertain about, given the absence of sector-specific regulatory guidance.

The implication is that AI's documented benefits, better care quality and stronger CQC outcomes, risk becoming available only to providers with the scale to navigate the market themselves. That is not a level playing field. And it is the kind of structural inequality that Andy Burnham, now installed as Prime Minister with a strong devolution brief and a long track record in social care reform, has consistently argued must be addressed at system level rather than left to market forces.

The Trademark Signal

The regulatory lag is visible in the intellectual property data too. According to AIBD analysis of IPO (TMD) data, Class 44 UK trademark filings, covering medical, veterinary and healthcare services, stood at just 803 in 2026-Q3, a 65% drop on the prior period. Whether that reflects post-peak consolidation among early-mover vendors or a broader hesitation about brand investment in an uncertain regulatory environment, the contraction is sharp. Vendors filing trademarks are signalling long-term commercial commitment; fewer filings may mean fewer new entrants betting on a stable homecare AI market.

What Regulators Need to Do

The Birdie report is direct about what it wants from regulators. It calls for guidance built specifically for homecare, not hospital-grade frameworks bolted onto a different care setting, and it wants inspectors equipped to assess how AI is actually governed in practice, not just whether a policy document exists in a filing cabinet. It also calls for support targeted at smaller providers who cannot access reliable information on their own.

None of that is unreasonable. The CQC has indicated it encourages innovation where it benefits people receiving care, but explicit inspector training on AI governance is not yet part of the standard inspection toolkit. The gap between what providers are doing and what regulators are equipped to assess is, at present, uncomfortably wide.

The frameworks will eventually catch up. By then, the technology will have moved again. That is almost certainly the point.