Accelerating justice decision-making with AI-powered guidance

Challenge

The Parole Board is responsible for conducting rigorous, fair, and timely risk assessments for individuals. Members of the Parole Board need to be well-informed about the most relevant laws, procedures, and guidance in order to make risk-based decisions. As such, they needed a quicker and more efficient way to access relevant content, as the existing manual process could be time-consuming. While an earlier prototype had been trialled, it fell short in performance, scope, and user experience - prompting a need for a more scalable and user-friendly solution.

The Parole Board sought a Retrieval-Augmented Generation (RAG) tool to streamline this process, reduce reliance on the internal policy team for routine queries, and ensure users always accessed the most up-to-date information.

Solution

Methods was engaged to deliver a two-month proof of concept (PoC) for an AI-powered policy retrieval assistant, designed to help Parole Board members get fast, accurate answers to policy questions. Working closely with stakeholders, Methods developed POLI - a secure, cloud-hosted chatbot integrated with SharePoint using AI Foundry and deployed in Azure. Key elements included:

• AI-ready architecture: designed and deployed scalable cloud infrastructure to support secure search and data processing.
• Automated indexing: creating a knowledge base that detects changes to SharePoint documents and updates the Azure AI Search index in real time, using Logic Apps, Python Function Apps, Microsoft Graph API, and Azure Event Grid.
• Interactive front-end: developed a clean, intuitive user interface for stakeholders to interact with the retrieval system. Designed a front-end UI in Python in Streamlit and deployed via Azure Web Apps via local Git integration to allow easy user interaction.
• Search capability: implemented a high-performance retrieval system to return contextually relevant policy guidance in response to user questions.

User acceptance testing, gathering feedback, and handover sessions ensured the solution was usable and maintainable for the Parole Board post-delivery.

Impact

The POLI chatbot was successfully delivered and adopted by the Parole Board, providing members with a responsive self-service platform for policy queries. This enabled:

• reduced pressure on policy teams by resolving first-line queries automatically
• faster, more reliable access to guidance, supporting timely and accurate assessments
• future-proofing: an architecture and indexing pipeline built for scale, ensuring users always interact with the latest available guidance
• positive user feedback: early adoption validated the solution's practicality and set the foundation for further enhancements.

The minimum viable product (MVP) not only demonstrated the solution’s immediate value but also established a roadmap and costed plan for potential future development.