UK social care gets first cross-sector AI guidance
UK departments issued the first cross-sector AI guidance for social care, citing a 63% administrative-time reduction in Swindon while keeping professional judgement central.

Why it matters
- November 20, 2026: The Department of Health and Social Care and the Department for Education issued joint national AI guidance for children’s and adult social care.
- The release is the first cross-sector national guidance of its kind, combining practical advice with governance, accountability, data protection and professional oversight.
- A Swindon pilot reported a 63% reduction in administrative time; typical Care Act assessment conversations fell from 1.5 hours to 30 minutes.
- Swindon assessment write-ups fell from four hours to 1.5 hours; Peterborough’s Hey Geraldine answered more than 1,200 questions and saved around 15 minutes per interaction.
- Four in five care providers use digital social care records, helping almost 90% of people who draw on care.
On November 20, 2026, the Department of Health and Social Care and the Department for Education issued joint national guidance for artificial intelligence in children’s and adult social care, citing a Swindon pilot with a reported 63% reduction in administrative time. The departments describe the release as the first cross-sector guidance of its kind.
Local authorities and care providers now have a common national reference point for deciding where AI fits in their work. The departments say shared expectations can make adoption more consistent and increase confidence across the sector.
Isabelle Trowler, Chief Social Worker at the Department for Education, said: “Organisations must take responsibility for governing, implementing and supporting its use.” That puts organisational responsibility at the centre of implementation rather than treating the purchase of a tool as the end of a deployment.
Takeaway: Local authorities and providers have shared expectations, while each organisation still makes its own adoption decisions.
What does the guidance cover?
The guidance gives leaders, practitioners and sector organisations a common national framework for adoption decisions. It combines practical advice with governance, accountability, data protection and professional oversight, and covers decisions by local authorities and providers.
Alongside that framework, the guidance names several information tasks that can sit inside existing casework:
- Transcribing and summarising case notes.
- Translating information.
- Searching complex case histories.
- Identifying gaps in support.
The stated purpose is to reduce administrative work so staff can spend more time supporting people and families. The departments also want services to become easier to understand and access.
These tasks operate on different parts of a case. Transcription and summaries condense case notes, while translation changes the language of information. History search and gap identification help practitioners find relevant material and identify missing support.
Takeaway: The guidance gives AI defined administrative jobs while leaving professional decisions with practitioners.
What did the Swindon pilot measure?
The pilot cited in the announcement took place within Swindon Borough Council’s adult social care work and used an AI tool to support social workers. The government reports that a typical Care Act assessment conversation shortened from 1.5 hours to 30 minutes.
Assessment write-ups fell from four hours to 1.5 hours. Staff with visual impairments and people for whom English was not their first language saw particular benefits from the tool.
The 63% headline therefore describes reported results from one local adult social care pilot, not a measured average for every provider. The reported savings cover two linked stages: the assessment conversation itself and the write-up that follows.
Takeaway: The Swindon pilot reports substantial time savings, but its results remain tied to one council’s adult social care work.
How does Peterborough’s Hey Geraldine work?
At Peterborough council, staff built a bespoke assistant called Hey Geraldine using the knowledge of an experienced occupational therapist. The system gave colleagues immediate access to guidance.
The government reports that Hey Geraldine answered more than 1,200 questions and saved around 15 minutes per interaction. Those figures measure interaction time, not assessment duration or case write-up time.
Peterborough’s result therefore answers a different question from Swindon’s: whether staff can retrieve specialist guidance quickly. The Swindon tool supports assessment work, while Hey Geraldine packages an occupational therapist’s expertise into an interface colleagues can query for immediate guidance.
Both projects give local teams a measurable starting point for judging wider use. They do not settle how a service should govern the technology, which is why the national framework assigns that responsibility to organisations.
Takeaway: Peterborough shows a second route to efficiency by giving staff immediate access to specialist guidance.
Why does professional judgement remain central?
Children and Families Minister Josh MacAlister said technology cannot substitute for a social worker’s knowledge and skill. He presented administrative relief as a way to preserve time for listening to children, working with families, and securing the support and protection they need.
That sequence matters because the proposed benefit reaches beyond lower paperwork. It aims to redirect time toward listening, family work and support decisions.
Trowler’s emphasis on governing, implementing and supporting use sets three organisational obligations:
- Govern the technology and its use.
- Implement it within the service.
- Support the people using it.
Those obligations make professional judgement a condition of adoption rather than an optional extra. A tool can reduce repetitive work, but the guidance leaves organisations responsible for how it fits into practice.
Takeaway: AI can save time, but people and organisations remain accountable for care decisions.
How does this fit wider digitisation?
The government reports that four in five care providers now use digital social care records. It says the records help almost 90% of people who draw on care.
The digitisation effort targets paper records, repeated conversations, errors and inaccuracies. The wider aim is for all care providers to become fully digitised by the end of the current Parliament.
The AI guidance therefore sits inside a broader shift in how care information is kept. Digital records can provide source material for transcription, summaries and history searches, while the framework requires organisations to govern data protection, implementation and practitioner support.
That connection makes the guidance relevant to more than AI purchases. It asks local authorities and providers to make adoption decisions with governance, accountability and professional oversight built in from the start.
Takeaway: AI adoption now overlaps with the move from paper to digital care records.
What happens after the guidance?
The practical next phase sits with local authorities and providers. They must turn the national principles into informed adoption decisions, using local measures of administrative benefit while retaining organisational accountability and practitioners’ professional judgement.
The Swindon project offers a measured benchmark for assessment and documentation time. The Peterborough project offers a different measure: questions answered and minutes saved per interaction with occupational therapy guidance.
The guidance treats both as local evidence rather than universal service standards. The next test is whether councils and providers can reproduce the reported time savings across different services without weakening human oversight.
If they can, AI in social care could move from isolated pilots toward the consistent, accountable adoption the departments want.
Takeaway: The pilots give councils measurable starting points, while accountability and professional judgement set the conditions for wider use.
Source: unite.ai
More from Sophie Lindqvist
Show full bio
Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.
823 articles
Related articles
- OpenAI unveils European youth safety blueprint and €500,000 in grants
- Australia to legislate mandatory safety rules for frontier AI
- AI agent governance needs three controls before autonomy expands
- HKUST's ASD Super Agent Hits 0.92 AUC in Early Autism Screening
- Australian AI governance overhaul proposed in Bond Law Review