Client profile
A UK software provider whose case management platform is used by fostering and adoption services to hold the full record for children in care: health plans, incident reports, risk assessments, placements and contact history. Their users are social workers under regulatory duty and time pressure.
The challenge
Case files are large and grow continuously. Social workers were spending too much time searching for key details, summarizing histories and checking that regulatory requirements were met. The bigger risk was what gets missed in that volume: an allergy buried in an old health plan, an incomplete risk assessment, a pattern of concern spread across several incident reports. The client wanted to help its users act faster and safer, but any solution had to process data entirely within their environment, respect existing user permissions, and produce answers that could be audited.
What we did
We designed an AI assistant that sits as an overlay inside the existing case file view, behind the client's existing single sign-on and multi-factor authentication. It only sees the case file the user currently has open, and only what that user is permitted to see.
The model is hosted within the client's environment, on-premises or private cloud. It is fine-tuned and grounded on the relevant regulatory framework, including the 2011 fostering regulations and national minimum standards, the platform's own data structures and representative case files, with prompt engineering focused on accuracy over fluency. Every interaction is logged for audit.
The work is phased:
- Phase 1, proof of concept (3 months). Contextual question and answer on the open case file (“what are the urgent matters for this child?”), a targeted set of regulatory checks and urgent risk surfacing (missing risk assessments, critical health concerns), data quality prompts for missing or inconsistent information, and a pilot with a selected group of social workers whose feedback shaped the next phase.
- Phase 2, proactive analytics. Automatic scanning of case files for emerging risks and patterns such as repeated names, missed assessments and behavioral trends, with notifications to the responsible worker, and extension of the assistant to the foster carer and young person apps.
- Phase 3, advanced features. Audio transcription and structured summarisation of recorded visits, cross-case pattern detection to spot common individuals or recurring incidents across multiple files, and integration with regulatory updates so the knowledge base stays current.
Fixed-price build, with an annual support and maintenance agreement covering security updates, minor enhancements and regulatory compliance updates.
What we delivered
- In-platform AI assistant with contextual Q&A on the open case file
- Regulation-grounded response layer aligned to fostering standards
- Data quality prompts flagging missing or inconsistent records
- Full interaction audit trail and role-based access enforcement
- Private, on-platform model deployment with no external data processing
- Pilot program design and feedback loop with practising social workers
- Phased roadmap for proactive analytics and cross-case pattern detection
Results
- AI assistant delivered inside the case file view within the 3-month proof of concept and piloted with practising social workers
- Contextual question and answer, regulatory checks and urgent risk surfacing in production use on live case files
- All processing on-platform: no case data leaves the client's environment, and every interaction is logged for audit
- Data quality prompts flagging missing risk assessments and inconsistent records at the point of entry
- Roadmap for proactive analytics, foster carer and young person app extension, and cross-case pattern detection agreed for subsequent phases