Agentic Systems in Government
Exploring what it means when AI agents, not forms, not chatbots, deliver government services at scale. A consolidation of my personal thinking on agentic government.
This project is designed to be a consolidation of my thinking about the topic of Agentic Government. The central question: what does it actually mean when an AI agent delivers a government service - not a better form, not a chatbot, but a genuine agent that understands intent and acts on behalf of a citizen?
Why this matters
Government service delivery is still largely built on the assumption that citizens will translate their own needs into bureaucratic language. A simple example is needing to fill out a complex form with little context or understanding.
The cost of that is enormous and mostly invisible. Agentic systems change that assumption fundamentally.
The policy and accountability frameworks haven’t caught up. This project is an attempt to think through what they need to look like.
What I’m working on
A key reference point is The Agentic State - a vision paper initiated and supported by the Berlin Global Government Technology Centre and the World Bank. I will use this to shape some of my thinking - as a strawman - about what I think could work, or the practical application of this in a government context. Alongside that I’m pulling together relevant academic work on accountability, human oversight, and the specific constraints of public-sector AI deployment.