AI should not make decisions about people, says new guidance on AI in social care in England
Councils may use AI to transcribe, draft and translate, but risk, need and eligibility calls stay with staff, under guidance published November 20.
By AI Desk ·
England's education and health departments on Friday, November 20, 2026, published guidance on how councils and care providers may use artificial intelligence in children's and adult social care. The government calls it the first cross-sector guidance on the subject. It is non-statutory: it sets out the two departments' position and leaves the decision to adopt AI with each organization, in line with its legal duties, data protection arrangements and professional standards.
The document is aimed mainly at local authorities, but the Department for Education (DfE) and the Department of Health and Social Care (DHSC) say its principles also apply to other organizations providing social care. It covers frontline practitioners, supervisors, data protection officers and procurement staff, among others.
Four principles and a list of no-go uses
The guidance rests on four principles. Casework stays the practitioner's responsibility, and whatever an AI tool produces is a first draft to be checked. Decisions that directly affect people, such as those on risk, need, eligibility, safeguarding, care planning or intervention, stay with practitioners rather than AI. AI use should leave people with better experiences and outcomes. And staff should put personal information only into tools their organization has approved, never into publicly accessible AI tools.
In a section on limits, the DfE says today's AI should be avoided or handled with significant extra caution in high-consequence work: deciding on its own who needs help or what intervention to make, writing straight into case records without meaningful human review, or assessing a person's needs or risks independently. Predictive tools that forecast an individual's needs, and agentic AI that completes tasks by itself, are listed as uses for caution. Children's safeguarding and statutory work should be treated as high-consequence by default, the governance checklist adds.
The document is blunt about error. "Hallucinations are a feature of AI tools, and not a bug," it says, and it lists bias, privacy and security risks and automation bias, where staff give AI output undue weight, as risks organizations should identify and manage. It also warns that a push toward digital tools can leave out people with little digital confidence or access.
What councils report
The government backs its case with results that councils reported from their own tools. The figures below are the councils' own results as the press release and the guidance's case studies report them.
| Council | Use | Reported result |
|---|---|---|
| Swindon Borough Council | AI tool supporting adult social workers | 63% less administrative time; Care Act assessment conversations down from 1.5 hours to 30 minutes, write-ups from 4 hours to 1.5 hours |
| Kingston Council | Drafting case notes, assessments and supervision records | 50% to 60% less time on case notes and assessments; supervision write-ups from 40 minutes to under 10 |
| Peterborough council | "Hey Geraldine" assistant built on an occupational therapist's knowledge | More than 1,200 questions answered, about 15 minutes saved per interaction |
The guidance also describes AI translation on a website for care leavers that offers content in more than 100 languages, while telling organizations to check accuracy and to consider a qualified interpreter where a translation feeds into a statutory process.
No technology can replace the knowledge and skill of a brilliant social worker, but by taking away some of the admin, AI can free them up to put those skills where they matter most.
Where the regulators stand
The guidance reprints the positions of the three regulators involved. Ofsted says it does not evaluate AI tools directly but its inspectors can consider the impact AI use has on children. The Care Quality Commission says it does not regulate individual AI tools, and it expects providers to keep human oversight of AI, be transparent about it and govern it well. Social Work England says social workers stay answerable for their decisions whether or not they use AI.
The departments call the document a foundation and expect further resources to follow. The press release also says four in five care providers now use digital social care records, and that the government wants every care provider fully digitized by the end of this Parliament.
Sources
- New AI guidance to cut paperwork and free more time for care Department of Health and Social Care and Department for Education (GOV.UK) gov.uk
- AI in social care (guidance) Department for Education and Department of Health and Social Care (GOV.UK) gov.uk
- Artificial intelligence (AI) in social care (publication page) Department for Education and Department of Health and Social Care (GOV.UK) gov.uk
Checked against the 3 sources above on November 21, 2026. Spotted an error? Tell us.