How Far Should AI Go in Public Sector Decision-Making?

How far should AI support decisions that affect people’s lives? Drawing on our work with Ulster University, we explore what the development of Hercules, an agentic AI proof of concept for the courts, can teach us about deploying AI responsibly in high-stakes public sector environments.

Our session will share practical lessons from developing the technology and engaging directly with members of the judiciary, including around trust, verification, transparency and effective human oversight.

We will also explore a more fundamental question: as AI systems become increasingly capable, where should the boundary lie between assisting human decision makers and exercising judgement itself? Attendees will gain practical insights to help their organisations assess the opportunities, limitations and governance considerations involved in adopting AI across public services.