What was the challenge?

While traditional recording systems are able to capture large volumes of social care case notes, they make it challenging to then identify patterns, surface lived experience or use narrative-based evidence to inform practice development and service design. Powys County Council sought to move beyond using AI solely for efficiency or administrative automation and instead explore how it could support service improvement in social care, while respecting professional judgement, ethical standards and coproduction principles.

Who would benefit and how?

The intention was to create a new participatory approach to social care that would put people’s lived experiences at the heart of service design, ensuring that their stories would actively inform learning and service improvement rather than being reduced to transactional records. 

In addition, social care practitioners would gain new tools to support reflective practice, supervision and professional development while, at an organisational level, Powys CC would be able to gather richer insights to help inform future service design and policy development.  

What was the solution?

In June 2025, Powys implemented Magic Notes as an AI assisted transcription and analysis system within social care, since when it has been used by more than 700 staff and processed over 28,000 records, generating time savings and enabling new forms of reflective and analytical practice.  

Building on this foundation, the council developed an advanced AI enabled narrative analysis approach which combined transcription with structured prompts and qualitative analysis techniques. A pilot project captured the personal stories of individuals with lived experience of substance use, with AI extracting transcripts, impressions and themes for human led validation and interpretation. The approach used a hybrid Labov–Riessman framework to support consistent, interpretable analysis while retaining narrative richness. 

Who was involved?

The project was led by the social care digital and development team, who used Participatory Action Research as its methodological foundation. This approach ensured that the research was co-produced with practitioners and people with lived experienced and that the findings would be robust and meaningful to professional and academic audiences. Indeed, coproduction was a core principle throughout the project, which included Social Care Wales and academic partners such as Manchester University, with feedback loops used to validate findings and ensure AI generated insights aligned with professional and lived experience perspectives.

What was achieved?

The use of Magic Notes enabled largescale capture and analysis of social care narratives, supporting improved reflective practice and learning at individual, team and organisational levels. The narrative analysis pilot demonstrated how AI could surface themes of identity, stigma, safety and systemic barriers that have previously been difficult to capture through traditional recording methods.  

Building further on the pilot, PCC are now exploring plans to apply the narrative analysis approach to thousands of Magic Notes case files, creating a rich and robust set of insights that can inform future practice development and broader discussions about service design and organisational change.  

How was the Welsh language considered?

The primary focus of the pilot was on how themes identified in lived experience narratives might inform reflective practice and, as none of the participants were Welsh speakers, the Welsh language was not an explicit consideration. However, the potential exists to extend the approach to include Welsh language case studies and materials where required and ensure future representation of the Welsh language. 

What were the key governance steps taken?

Practitioners were supported through the development of prompt ‘playbooks’ which guide users through structured AI analysis while maintaining transparency and human insight at every stage. These playbooks are designed to support learning and adoption rather than automate decision making, an approach that encourages iterative skills improvements and reinforces ethical safeguards, thus helping to embed AI as a support to professional experience rather than a standalone tool.  

Strong emphasis was placed on human in the loop oversight and responsible AI engineering. Adhering to principles such as consistency, repeatability, reversibility, interpretability and human accountability, AI outputs were continuously reviewed and interpreted by practitioners, with professional judgement retained at all stages. This journey, with iterative testing and human review built into the workflow, enabled frequent opportunities for checks around safety, usefulness and alignment with professional values, helping to make any improvements deliberate, slow and visible while ensuring that AI supports rather than replaces professional decision making. 

What lessons have been learned?

General lessons included the importance of moving slowly and deliberately when applying AI in sensitive service contexts, maintaining visibility of how AI is used and grounding innovation in coproduction and lived experience. The project also demonstrated that AI can be used not only to improve efficiencies but also to challenge an organisation’s ‘muscle memory’ of entrenched behaviours and structures, disrupting established patterns and introducing new, more adaptable perspectives. 

More specifically, the outcomes of the project demonstrably have the potential to be scaled nationally, positively influencing policy and practice across the Welsh public sector and beyond. 

Supporting information

Learn more about the Participatory Action Research methodology here 

Discover more about Social Care Wales’ thoughts on relationship-centred care here