A nationwide mapping of agentic AI readiness in the Finnish public sector is here – Sitra suggests what should be done next

Public-sector organisations across Finland – and around the world – are running AI pilots by the dozen. Yet pilots focusing on automating current tasks are not adding up to real renewal. One of the reasons, Sitra and The Agentic State foundation argue, is that a crucial level of analysis has been missing from the approach: the function.

The function refers to a public-sector activity with a clear result that often recur in various organisations, such as receiving an application, assessing eligibility, issuing a licence or making a payment.

Today, AI initiatives are framed almost entirely around organisations, existing services and individual systems. With that framing it remains hidden how much work is actually duplicated. A ministry, a wellbeing services county and a municipality may all be redesigning the process around application intake in parallel, unaware that they are solving the same problem.

“The right question to ask in public organisations is not which task we should automate or which AI tool to adopt. We should first ask, on which level public-sector work should be understood, compared and redesigned”, says Jonna Heliskoski, Programme Director at Sitra

Sitra and the Agentic State foundation propose the function-approach as a fix: It means looking at common public-sector functions, as for example issuing a license, as the unit for renewal. The function-approach has also been utilized by the World Economic Forum, which has categorised the work of public bodies into 70 functions.

“Looking at the function level helps to make the repeating work visible”, continues Heliskoski.

Why now? Agentic AI changes what’s possible

The timing of Sitra’s working paper and The Agentic State foundation’s assessment report, is tied directly to a shift in AI capability. Unlike tools built to support a single task, AI agents can increasingly carry out bounded, multi-step work across systems and organisational boundaries.

That capability makes it possible to redesign recurring public-sector work as a whole, rather than simply bolting new technology onto existing organisational processes.

The evidence base: The world’s first nation-level mapping exercise

The proposition to advance the function-based approach in Finland is grounded in wider research. First Moves: A Finland-Specific Assessment of Agentic AI Potential in Government is the first study in the world to map the potential of agentic AI across a national public sector using a functions-based approach. The report, produced by The Agentic State foundation in collaboration with Sitra, maps where recurring public work may be technically and legally suitable for agentic AI.

“Finland is the first country in which this function-based approach has been tested at national level. We mapped 70 recurring government functions across 109 Finnish public sector organisations and scored each function’s readiness for agentic AI”, elaborates Luukas Ilves, a co-founder of the Agentic State foundation and one of the authors of the report.

The mapping is, however, a starting point. It does not capture the workflows, costs, benefits or readiness of individual organisations. More practical work of deciding what to prioritise in adoption of agentic AI still lies ahead. That is the gap Sitra’s working paper The Next Questions on the Path to the Agentic State begins to fill. It lays down the relevant questions to be answered next in order to advance the function-based approach in public-sector renewal in Finland.

What Finland should do next: A cross-organisational test with one recurring function such as application intake

Sitra proposes that the next step for the Finnish public sector should be to choose one function and run a test with multiple organisations involved.

For this first test, a strong candidate would be application intake. The function includes steps such as identifying the applicant, requesting missing information, and preparing the case for human review.

In the test, participating organisations should establish the baseline, redesign the workflow and data before procuring technology. It should be tested whether a reusable AI capability, partly shared across organisations, can work for several public bodies.

“Find one important function that several public bodies already perform, make the shared work visible, and decide what those public bodies should redesign or build together,” concludes Sitra’s Senior Lead Antti Harjunpää.