Transforming adult social care delivery with generative artificial intelligence

Through the innovative adoption of Generative Artificial Intelligence tools, Neath Port Talbot Council has been able to transform the way our Adult Social Care team engage with our residents.

It is well documented that all local authorities are under extreme financial pressure at a time when service demand in key areas is increasing exponentially. Neath Port Talbot Council is no exception.

Adult Social Care is a service area identified as experiencing 
extreme delivery pressures.  Challenges around the recruitment and retention of social workers, at a time when there is increased demand for various reasons, including an ageing population and residents presenting with more complex needs, are resulting in a significant strain on an already stretched workforce.

To address these issues, the council is integrating Generative Artificial Intelligence (Gen AI) into operations. It became clear that social workers spend a considerable amount of time on administrative tasks and back-office engagements.
Through engagement with our wider local government networks, we identified Magic Notes, an AI transcription tool specifically designed for social workers that generates detailed assessments from an audio recording, which had some impressive claims regarding time savings and service improvements.

We were the first local authority in Wales to engage with the supplier, agreeing on an initial three-month pilot of the technology for our adult social care team. A project team was established within the service area to complete the necessary Information Governance requirements and deal with the contract.

At the outset of the pilot, we identified three key metrics to test the product against:
1. Time savings - reduced time spent on admin tasks and meeting write-ups.
2. Quality and Accuracy – Work submitted should be of high quality, with a transcription accuracy rate of over 90%.
3. User sentiment - Social workers involved in the pilot adopt and enjoy using the tool.

Time Savings
Practitioners saved an average of 25 minutes per assessment write-up, totalling about 7.5 hours a week each. Assessments were submitted 8 hours faster with the product, and all participants reported significant time savings from the technology.

Quality and Accuracy
Throughout the pilot, practitioners remarked on the impressive accuracy of transcriptions produced and found them to be 97% + accurate, far higher than other products that had previously been tested, which built further user confidence in the technology. Feedback also showed that the quality of the conversations whilst using the product was significantly higher than when carrying out traditional ‘pen and paper’ assessments, as it allowed a more open, flowing conversation with clients.

User sentiment
Given the significant time savings and high levels of accuracy, the practitioners involved unanimously endorsed the product and wanted to continue using it. Managers commented on the real headway being made into the backlog of cases, as well as being able to rebalance the work allocation across the workforce, all of which had a positive effect on the health and well-being of our social workers. Additionally, users were invited to share feedback throughout the pilot and discuss their experiences with the project team, allowing for learnings to be fed directly into the product roadmap and user support plans.

The pilot process required engagement with a wide range of stakeholders, including practitioners, councillors, trade unions, and the wider workforce, which was conducted through webinars, product demonstrations, and in-person workshops. This ensured that the project team established a shared understanding of the pilot’s scope and provided communication channels where any concerns could be openly discussed and addressed.

Challenges and considerations
Despite the success of the pilot, the adoption of new technology and methods of working is never without challenges. Staff experienced occasional technical issues with the product, such as connection outages and the need to keep the recording device battery charged. And although the transcriptions were found to be accurate, the need for proofreading remains.

We also found that the use of Magic Notes was most effective for experienced staff – manual notetaking is an important foundational skill for new starters in terms of building knowledge, over-reliance on technology, with all the potential issues that can occur, is something managers must be aware of and consider in training plans. Finally, ethical considerations such as data protection and responsible use are essential.

Outcomes, Future Targets, and Goals
The use of Gen AI tools across service areas in Neath Port Talbot Council continues to grow from strength to strength. Through our investment in additional capacity to support the user base, provide training, and develop case studies, we’re seeing the understanding of how these tools can be successfully deployed to support service redesign gain significant traction.

In our social services directorate alone, given the success in adult services, we’re already developing specific pilot projects with the children’s, housing, and community teams, where they anticipate significant time savings and service enhancements using these technologies.

We’ve already undertaken extensive engagement with other local authorities in Wales, not only sharing our approach and pilot findings, but also helping give colleagues a head start by openly sharing our data protection impact assessment, solution assessment, standard operating procedures, and more, allowing them to expedite their consideration of the product.  This has led to ten local authorities in Wales now signing up to their own pilots.

Our experiences of Gen AI are already demonstrating the fundamental impact that the use of these technologies is having on our ability to innovate service delivery.  To meet the pressures currently faced in local government, we must continue embracing these technologies and develop use cases to showcase the art of the possible, all while balancing and managing the risks presented through appropriate governance frameworks, continuous monitoring, and a commitment to ethical AI practices.