Derby City Council’s use of AI to assess people’s care needs splits opinion. Some see a way to stretch scarce resources; others fear breaches of rights and new inequities. Used on the right problems, with human-rights safeguards, AI could help councils improve efficiency and outcomes.
London councils use AI dashboards combining provider capacity, quality and local context to help commissioners spot gaps, model scenarios and make quicker, more consistent decisions. Shared with providers, they allow errors to be challenged and local insight to be added – supporting fairer contracting and resilient markets. Councils must use dashboards alongside a fair-cost floor, not as price levers.
Transcription and summarising tools can turn meetings and case discussions into minutes, action lists and summaries. Used safely, they restore time for direct work with people.
Pilots are exploring prompts and monitoring to help people stay independent, flagging concerns for early intervention. Norfolk County Council uses predictive analytics to identify people at risk of falls. These tools add value when targeted, explainable and part of an agreed care plan, with ways to pause if unsuitable for the person.
Pattern-spotting models can highlight safeguarding cases for human review, but they must support – not replace – professional judgement, with clear routes for challenge and override.
The risk is real. Historic data encodes inequities; without testing and monitoring for bias, systems will reproduce them, especially in eligibility and safeguarding decisions. Councils must start with equality impact assessments and ‘shadow’ pilots against historical cases. Opaque ‘black box’ outputs are hard to justify to practitioners, families or the public; in high-stakes pathways, inscrutability is unacceptable.
Councils remain responsible for personal data collected and used. They should complete data protection impact assessments (DPIA); minimise data shared with vendors; and write contracts requiring transparency on training data, model updates and subcontractors. Councils need access to logs and performance reports, bias-testing regimes and a quick way to suspend a tool if issues arise. They need to publish plain-English DPIA summaries and model cards for public scrutiny.
Staff members also need confidence. Many staff experiment with public chatbots to save time. Without employer-provided alternatives and clear policies, the DIY approach risks confidentiality breaches and inconsistent practice. Employees want approved tools and everyday training. Some councils are responding with mandatory training before access to AI and by building digital-champion networks.
Co-production with staff and people with lived experience ensures technology serves, not replaces, human relationships. Swindon Borough Council’s work with people with learning disabilities in developing AI-powered easy-read tools shows the value of inclusive design.
Councils must:
- Start where risk is low and value high: Minutes, summaries and drafting templates, plus shared commissioning dashboards.
- Measure outcomes openly and independently, including perceived fairness and time returned to practice.
- Keep professionals as final decision-makers where eligibility, risk or rights are at stake – and record when a tool was used, what it suggested and why a decision followed or departed.
- Choose explainable systems by default; if a tool cannot be justified in plain language to the person and family, do not use it.
AI is not a silver bullet, but it can be a useful spanner. If councils aim at administrative friction and market visibility first; insist on human oversight, scrutiny and strong data protection; and co-produce with the people affected, AI can help adult social care deliver more time for relationships, better decisions and a fairer, more resilient system.
Dr Jane Townson OBE is Chief Executive Officer at the Homecare Association. Email: [email protected] X: @homecareassn