Search across all content
Three councils jointly rolled out a commercial generative AI tool to cut administrative time in social care and customer service, sharing learning across the three.
In the United Kingdom, councils are the units of elected local government that deliver many of the services people rely on, among them social care, the publicly funded support provided to vulnerable adults and children. Much of that work depends on time spent with people in person: visiting them at home, talking through a problem, judging what kind of help they need. Every one of those encounters also generates paperwork. Workers write up what happened, go back over earlier case files, and complete the formal assessments that set out a person's needs and any risks to their safety, the documents that determine what support they receive.
For councils in South Wales, that administrative load was taking up time that staff would otherwise spend with residents. Social care, in particular, is a high-pressure field, and councils have seen experienced staff leave the profession, making recruitment and continuity of care harder to sustain. Customer-service teams faced a related strain: when a resident called with a query, agents often had to search through lengthy policy documents or council web pages while the caller waited, which slowed responses and made it harder to guarantee consistent answers.
To return some of that time to residents and casework, the three councils adopted Microsoft 365 Copilot, a commercial generative AI assistant built into the Microsoft Office software the councils already used. Generative AI of this kind produces text, such as a draft or a summary, in response to written instructions known as prompts. Swansea, Carmarthenshire and Rhondda Cynon Taf chose to adopt the tool together, sharing what worked and running joint skills-building events so that the effort spent solving a problem in one council could benefit the others.
Each council applied it to its own priorities. In Rhondda Cynon Taf, social care staff use it to draft and summarise documentation for assessments and case notes, with the worker retaining responsibility for the final record. Swansea rolled it out across housing, education, customer service and social care, and gave staff access to its conversational version alongside a set of purchased licences. Carmarthenshire focused on customer service, building a chatbot, an automated question-answering tool, that draws on the council's own policies and documents after they were moved into a central store on SharePoint, Microsoft's document platform, so contact-centre agents can find the right answer while a resident is on the line.
1. Social care staff report spending less time on paperwork
At Rhondda Cynon Taf, where the council reports that around 350 staff use the tool, needs assessments are said to be completed nearly four times as quickly, case-note reviews are more than twice as fast, and nearly twice as many risk assessments are completed within an hour. Mari Ropstad, who leads adult social care there, ties this to wellbeing and retention, summarising the logic as "happy staff stay". The council reports that 86% of staff feel the quality of their work has improved, 75% feel more positive about their role, and 50% report better wellbeing.
2. Thousands of hours saved
Swansea, with 2,500 staff regularly using the conversational tool, estimates it saved 5,400 staff hours in a single four-week period, which its corporate-services director, Ness Young, describes as time redirected into housing casework, social care and customer service.
3. Customer-service responses became faster
In Carmarthenshire, the chatbot has helped contact-centre staff answer residents more quickly, and the council reports that complaint response times now fall within its ten-day target. Staff who are neurodivergent, particularly those with ADHD or dyslexia, report that the writing and summarising features make clerical work less overwhelming.
Confidence and digital skills proved to be the main obstacles to adoption. The councils found that some staff were less comfortable with digital tools than expected, with ITPro reporting that Swansea's Ness Young described digital literacy as an obstacle to early use, one that was met through heavy investment in training, a support partner, and one-to-one help. Rhondda Cynon Taf found that staff needed concrete use cases and ready-made prompts in addition to general training and used experienced practitioners to coach their peers towards a consistent standard.
A phased rollout through willing early users supported adoption. All three councils began with a small group of more confident, enthusiastic staff, sometimes called hero users, who tested the tool and later passed prompts and methods to colleagues. Swansea issued its first 100 licences, mostly to pilot teams and partly to senior leaders, so the effort was visibly led from the top.
In a sensitive service, staff concerns had to be addressed directly. At Rhondda Cynon Taf, early hesitation centred on the ethics of using AI in assessments and on whether it might affect social workers' professional registration. The council worked through team-level volunteers and reassured staff that their judgement and professional expertise remained at the centre of work.
The reusable part is the adoption approach. Because the technology is a commercial product available to any organisation already running Microsoft's office software, the part another government could take from this is the way it was introduced: collaborating across institutions to share learning, phasing the rollout through early adopters, investing in confidence and skills as much as in access, and working through staff and ethical concerns before scaling.
Launch year: 2026
This case study was written with assistance from artificial intelligence.





Connect with 500,000+ public servants solving your hardest challenges.





Connect with 500,000+ public servants solving your hardest challenges.
Help public servants worldwide learn from your work, what worked, what flopped and what you'd do differently
Share your project
Log in or sign up to continue the conversation