The video features William Tarr and Louis Hogarth from the UK Ministry of Justice’s Justice AI unit, discussing their innovative approach to deploying AI products within government. William, a forward deployed engineer, explains that their team operates like a startup embedded within the government, focusing on building practical AI tools that frontline staff actually want and use, rather than traditional bureaucratic pilots and presentations. He highlights the challenges in government systems, such as disjointed legacy technology, poor organization, and a disconnect between policymakers and frontline workers, which their team aims to bridge by working directly with prison officers, probation officers, and courts.
Louis shares his experience working closely with probation officers, emphasizing the importance of understanding their pain points and building solutions rapidly based on real user feedback. Their team operates with a lean structure, often building minimum viable products (MVPs) within weeks and rolling out pilots nationally within months—an unusually fast pace for government projects. They spend significant time on-site with users, iterating quickly and prioritizing practical impact over theoretical discussions, which helps them navigate government bureaucracy more effectively.
Both speakers describe the unique challenges of working in environments like prisons, where infrastructure limitations such as lack of Wi-Fi and outdated hardware complicate technology deployment. They adapt by designing offline-capable solutions and maintaining close communication with frontline staff to understand and solve these problems in real time. This hands-on approach allows them to respond swiftly to issues and continuously improve their products, fostering trust and collaboration with users who often have little prior exposure to AI tools.
The team also addresses governance and security concerns by integrating best practices from day one, engaging with unions, ethics colleagues, and policy teams to ensure transparency and safety. Despite resistance and mistrust of AI within some parts of government, their embedded model and focus on building tools that genuinely help staff have enabled them to make significant progress. They also look ahead to the future of justice, exploring advanced AI applications such as CCTV analytics and predictive intelligence to further empower frontline workers and reduce administrative burdens.
In closing, William invites interested engineers to join their growing team, emphasizing the success of their forward deployed model and the impact they have achieved in a short time. They stress the importance of building with users, keeping solutions simple and accessible, and maintaining a problem-first mindset. The session ends with a Q&A addressing practical issues like infrastructure challenges and user education, reinforcing the team’s commitment to hands-on, user-centered innovation within the complex government environment.
