Lancashire County Council has officially announced that its social care workers are now using Microsoft 365 Copilot to draft structured case records from spoken visit notes. The move, revealed on June 30, 2026, marks one of the most concrete applications of generative AI in UK local government to date. The council estimates the new workflow will save thousands of hours of administrative work each year, allowing social workers to spend more time with vulnerable adults and children.

Social workers routinely record their observations after home visits, a process that can take up to an hour per case. By enabling them to dictate notes directly into Microsoft Word or Teams, then letting Copilot summarise and format that audio into a pre‑defined template, the council is cutting that time dramatically. Early internal tests showed that the time from dictation to a completed case note fell from around 45 minutes to under five minutes.

The initiative is part of a wider AI adoption programme within the council, which has been piloting Microsoft 365 Copilot across several departments since early 2026. Lancashire is among the first local authorities in England to deploy generative AI in a frontline social care setting, and its experience is being watched closely by other councils grappling with severe staff shortages and rising case loads.

How the system works

The technical workflow is deceptively simple. A social worker visits a client, and on returning to their car or office they open the Microsoft 365 mobile app on a council‑issued smartphone. They tap a button to start recording, speak their observations – describing the client’s condition, home environment, any safeguarding concerns, and next steps – and then tap stop. The audio is securely uploaded to the council’s Microsoft 365 tenant, where Copilot transcribes it using the same speech‑to‑text engine that powers Teams live captions.

Once the transcription is complete, Copilot applies a series of prompts designed by the council’s digital transformation team. It identifies key pieces of information: the client’s name and date of birth, the date and time of the visit, a summary of the conversation, any risk flags, and actionable follow‑up tasks. That output is then inserted into a Word document that has been pre‑configured to match the council’s case management standards. The social worker reviews the draft, makes any necessary edits, and saves it into the adult social care case management system, which in Lancashire is Liquidlogic’s Adults’ Social Care System (LAS).

Because the entire process happens within the council’s controlled Microsoft 365 environment, none of the data leaves the UK data centres that Microsoft guarantees under its public sector cloud contract. The council’s data protection team has signed off on the approach, noting that the audio recordings are automatically deleted after 30 days, unless a manager specifically preserves one for audit or supervision purposes.

The productivity dividend

Lancashire County Council employs around 800 social workers and social care assessors across its adult and children’s services. On average, each of them completes between four and six home visits a day, generating at least one case note per visit. Internal analysis found that social workers spent roughly 12 hours a week on documentation alone.

By shifting to AI‑assisted note‑taking, the council projects that productivity gains will free up the equivalent of 50 full‑time social work posts, without reducing headcount. Instead, those hours can be redirected to direct client contact, complex case planning, and supervision of newly qualified staff. The council’s director of adult social services, speaking at a regional digital health conference earlier this month, described the early results as “transformational”.

“What we’re hearing from practitioners is a real sense of relief,” she told delegates. “The tool isn’t replacing professional judgement – it’s removing the drudgery of typing up notes. One experienced social worker told me she used to dictate her notes into a separate dictation machine, then type them up at home in the evening. Now she leaves at 5pm and her notes are already in the system.”

Training and trust

Introducing AI into a profession that deals with highly sensitive personal data was never going to be a simple switch‑flip. Lancashire’s rollout followed a six‑month co‑design phase that involved social workers, team managers, legal advisers, and the council’s information governance board. The most critical step was building a set of custom Copilot prompts that correctly interpreted the informal language social workers use in spoken notes.

“A social worker might say ‘Mum was a bit down today, the house was untidier than usual, and I’m worried the medication is being missed.’ Copilot has to understand that ‘mum’ refers to the client’s mother, that ‘down’ is a clinical observation, and that medication non‑adherence is a risk flag,” explained the council’s head of digital innovation.

To build that understanding, the team ran hundreds of anonymised historic case note samples through the prompt chain, refining the output until it matched the quality of a manually written note 98 percent of the time. Participating social workers then received two half‑day training sessions, plus a weekly “AI surgery” where they could drop in and ask questions.

Uptake has been high. Three months after the first wave went live, 73 percent of eligible adult services staff were using the software at least once a day. The council plans to extend the tool to children’s social care by the end of the year, though the prompts will need significant re‑engineering to account for the different legal frameworks and recording standards.

The wider context: AI in local government

Lancashire’s announcement comes as councils across the UK face a £4 billion funding gap, according to the Local Government Association. At the same time, social care demand is rising sharply, with an ageing population and increasing numbers of children entering care. The government’s recent white paper on AI in public services explicitly encouraged local authorities to explore generative AI for administrative tasks, but few have moved as fast as Lancashire.

Other councils have experimented with Microsoft Copilot for generic office work – drafting emails, summarising meetings, writing reports – but the application to social care case notes is believed to be a UK first. The council is sharing its prompt libraries and governance framework through the national Local Digital programme, so that other authorities can replicate the model without having to start from scratch.

Microsoft itself has been actively promoting Copilot’s potential in frontline public service. A spokesperson for the company said: “We’re delighted to see Lancashire County Council pushing the boundaries of what’s possible. When you combine the security and compliance of Microsoft 365 with the power of large language models, you can transform services without compromising safety or privacy.”

The boundary between assistive and autonomous

Despite the enthusiasm, the rollout has not been entirely without friction. Some social workers initially feared that the AI would eventually replace their professional judgement, or that it would be used to monitor their productivity. The council has moved to address those concerns by making the AI assistance opt‑in – no social worker is required to use it – and by emphasising that the final clinical and legal responsibility always rests with the human professional.

Union representatives were consulted early in the process. Unison, which represents most council social workers in England, has generally taken a cautious but supportive line. Its national officer for local government said: “If AI can genuinely cut bureaucracy and let social workers get back to the work they trained for, that’s a good thing. But it has to be introduced in a fair and transparent way, and we’ll be monitoring Lancashire’s project closely.”

The biggest technical challenge so far has been handling regional accents and dialects. Lancashire’s social work workforce reflects the county’s diverse communities, and early versions of the transcription engine occasionally stumbled over strong local accents or non‑standard English. Microsoft has since rolled out a custom speech model that was fine‑tuned on a dataset of Lancashire‑accented voices, collected with consent from staff volunteers. The model now achieves over 95 percent word accuracy in internal tests.

Data protection and ethical governance

Any system that processes personal data about vulnerable individuals must meet the highest standards of the UK GDPR. The council published a data protection impact assessment (DPIA) alongside the launch, which outlines the controls in place. The audio recordings are processed in real time and are not used to train any AI model outside of the council’s tenant. The transcriptions are treated as case records and are subject to the same retention periods as manually typed notes.

A particularly sensitive issue is the possibility that a social worker might inadvertently record a third party – a family member, a neighbour, or a child – during a dictation. The DPIA addresses this by mandating that dictation must be done in a private space, and by programming Copilot to strip any names it does not recognise as a client or known associate before inserting the note into the case file. If the system does capture a third‑party name, it is flagged for manual review.

The council’s Caldicott Guardian, who oversees the use of confidential patient and service‑user information, signed off on the approach after an independent audit confirmed that the additional risk was minimal. “The alternative is social workers scribbling notes in their cars, then typing them up hours later,” the Guardian said. “This system is actually more controlled, because the audit trail is complete and the data never rests on an insecure device.”

What comes next

Looking ahead, Lancashire County Council is already exploring additional uses for Copilot within social care. One pilot in the early planning stage would use the tool to scan historical case files and surface potential patterns, helping team managers identify cases that might have drifted or that need a higher level of intervention. Another would integrate Copilot with the council’s finance system to automatically generate cost reports when a care package changes.

The council is also working with Microsoft to develop a version of the tool that can work offline. Social workers in remote parts of the county, such as the Forest of Bowland, often lose mobile signal during home visits. Storing the audio locally on the device and syncing it when the worker returns to range would resolve that problem, but it introduces new security challenges that the council’s IT team is still working through.

For now, the mood in Lancashire is cautiously optimistic. The technology has passed its first real‑world test, and the early data suggests it is delivering on its promise to give social workers back their most precious resource: time. As one frontline practitioner put it during a feedback session captured by the council’s internal communications team: “I went into social work to help people, not to write essays. This finally feels like a tool that’s on my side.”