10.06.2026 at 06.45
Performance audit

Use of artificial intelligence in central government

The central government and wellbeing services counties see plenty of potential in the use of artificial intelligence. While the authorities have a clear need to develop joint AI solutions, the central government lacks an operating model and structures to support this. Legislative barriers should additionally be dismantled.

Photo: Getty Images

Download report: NAOF's conclusions and recommendations: Use of artificial intelligence in central government PDF 191 KB

Scope of the audit

The preconditions of central government and the wellbeing services counties to use AI in their activities.
The framework and boundary conditions for using AI.
The situation and expected benefits of AI use.
Common operating models that promote AI introduction.

Key findings

Whereas many types of AI-related development are under way in central government, a clear overview and coordination are lacking.
The authorities develop overlapping solutions, which is neither sensible nor cost-effective overall.
Legislation and different interpretations associated with it are a key issue that hampers and slows down AI use.
Support for AI use available for the authorities is scattered, and they do not find it sufficient.
The authorities have many AI-related networks that are important for sharing lessons learned and experiences.

Briefly

The topic is current and important, and AI use has significant potential to improve the efficiency and services of central government.

Almost all organisations that participated in the audit have identified use cases for AI. However, a common challenge lies in how to move forward from experiments and achieve continuous activities.

Narrowly focused development should be replaced by an overall reform of processes, as individual experiments will not add up to a productivity leap.

There is a major need for cooperation and common solutions in the development and use of AI. While the authorities are willing to work together, the central government lacks an operating model and structures that would support this.

AI use is limited by interpretations of legislation on what tasks and data may be entrusted to AI. Whereas some legislative barriers have already been identified and certain ones also dismantled, much remains to be done in this area. In addition, information steering should be more uniform and serve the authorities’ needs better.

83% of the respondents find that AI will have significant or very significant potential benefits. 17% believe that the potential benefits will be limited or moderately significant. — Authorities’ views of potential benefits of AI over a time span of 3 to 5 years. Source: Survey conducted for the audit (n = 75).

61% of the respondents find that the current legislation does not enable efficient AI use in their organisation’s work. 84% have identified legislative factors that prevent or restrict AI use. — Authorities’ views of statements concerning legislation in the audit survey. Source: Survey conducted for the audit (n = 75).

According to survey findings,

90%

of organisations have striven to develop their personnel’s AI competence.

According to survey findings,

85%

of organisations have used other authorities’ lessons learned and experiences in their AI development.

Recommendations of the National Audit Office

The Ministry of Transport and Communications and the Ministry of Finance should develop the situational picture available on AI development in central government to meet the needs and purposes of different target groups.

The Ministry of Finance should define operating models and structures that support the development and deployment of authorities' shared AI solutions.

Together with the other ministries, the Ministry of Finance should continue to dismantle legislative obstacles to digitalisation with a more determined approach and ensure that legislation with essential impact on the authorities' work is compatible with AI use.

The Ministry of Finance should promote more consistent, needs-based and easy-to-use information steering related to AI use.