Stop Starting with Ministries: Mapping the 70 Functions of Government

It has been predicted that over 40% of agentic AI projects will be canceled by 2027 due to escalating costs, unclear business value, or inadequate risk controls — and situation won't be different in the public sector.

A new study suggests that this is not because the technology is weak, but because of structural misalignment: agentic AI operates across end-to-end workflows, while governments organise around ministry silos.

The Agentic State co-author Manuel Kilian — lead author of the study — mapped a way out of this mismatch. Instead of asking “how can ministry X adopt AI?”, governments should start from what they actually do. The map identifies 70 core functions of government — spanning service delivery, policy and governance, and enabling infrastructure — and plots each on a “topography of potential” across AI potential and implementation complexity, showing where agentic AI is ready to take on real work first. Often, the biggest wins are in the back office.

The map was launched by the Global Government Centre of Berlin with World Economic Forum and Capgemini, and you can find here a presentation of the findings, including a recorded conversation on them. Read about Manuel’s presentation of the study on Govtech 4 Impact World Congress blog.