Agentic AI, which has taken the form of systems that combine the intelligence of LLMs with the ability to act in real-world systems, marks a step change for government operations. Today, most governments are investing in it without a systematic view of where it fits. That gap produces two failure modes: valuable applications go unexplored while attention concentrates on visible but low-impact use cases, and deployments proceed into work that is unsuited or environments that are unready, producing costly failures that erode trust and political capital. So, how should governments get started with agentic AI?
First Moves: A Finland-Specific Assessment of Agentic AI Potential in Government answers that question for one country in full. It builds directly on the 70-function government taxonomy set out in Making Agentic AI Work for Government (World Economic Forum, Global Government Technology Centre Berlin and Capgemini, 2026), applying it across the whole of Finnish public administration — 69 central-government entities, all 21 wellbeing services counties, and the 19 region capital municipalities. Each function is scored on six dimensions covering capability, impact, productivity measurability, and technical, organisational and legal readiness, grounded in over 23,000 verified passages drawn from the documents public bodies publish about their own work.
The result is a data-driven portfolio of action paths. In central government, 14 functions offer a warranted deployment case now, concentrated in the financial backbone, the citizen-facing front line and enterprise support. Sixteen are ‘agentifiable’ and measurable but blocked — none of them by technology alone, which redirects the intervention from infrastructure spending to specific organisational and legal conditions. Six should remain human-led, because their core is negotiating between conflicting interests and carrying accountability for contested decisions. Applying the same framework to the counties and municipalities surfaced important differences: the same function can be a priority at the central level but constrained locally, so there is no single nationwide deployment list.
The findings are Finland-specific, but the method has a capability to be reused by any administration. Read the key findings or the full report. Navigate on the interactive dashboard to explore all 70 functions, their scores, the reasoning behind each one and the underlying evidence. Read Sitra’s working paper identifying the next right questions on the path to the Agentic State.
