In November 2024, Sweden's AI Commission delivered its roadmap earlier than planned. That was in itself a signal: development is fast enough that the normal pace does not suffice.[1]
Since then, the government has taken important steps with a national AI strategy and an action plan. But what continues to be underestimated in Swedish debate is not just the importance of AI, but its character. This is not a narrow innovation question. It is simultaneously a labour market question, an administration question, an education question, a competitiveness question and ultimately also a preparedness question.[2]
Three recurring mistakes
Pace error: policy often works as if change were slow and even. But within twelve to eighteen months capacity, costs and usage patterns can shift substantially.
Scope error: AI is often placed in a narrow compartment — research, innovation or digitalisation — despite effects cutting across multiple systems simultaneously.
Responsibility error: many see the problem, but too few have the mandate to prioritise across departments, agencies and sectors. The result is activity without sufficient force in implementation.
Why normal process is not enough
The democratic process should not be replaced. Inquiries, consultations and the rule of law exist for good reasons. But when an area moves quickly, the state needs to be able to complement normal process with shorter decision loops, clearer prioritisation and more operational coordination.
Otherwise a peculiar situation emerges: society produces well-written strategies whilst organisations in practice are already forced to improvise. That gap is expensive. It leads both to missed opportunities and to worse risk management.
What a more proportionate response would look like
First, a small permanent AI coordination unit is needed with the mandate to coordinate across labour market, education, administration, security and business policy. The question is too cross-cutting to be left to informal cooperation.
Then an action programme at the shop-floor level is needed. The state should identify a limited number of high-priority use cases in the public sector where AI under human oversight can deliver clear effect, and an equally limited number of areas where risks require particularly tight control. Today general ambition is often mixed with unclear prioritisation.
Finally, the transition track needs to be faster. It is not just about more training places, but about modular skills development, shorter learning loops and earlier signals to occupational groups whose work content changes rapidly.
What should be built in the first hundred days
1. A coherent situational picture. Which occupational groups, processes and public sector functions are affected first? Where are the biggest efficiency gains and where are the biggest risks?
2. A plan for data, procurement and responsibility. Many AI initiatives fail not on model quality, but on data, law, security and responsibility allocation not being solved in practice.
3. A fast track for public sector competence. Managers, lawyers, buyers, development specialists and union representatives do not need the same training, but they need a common language and faster access to relevant knowledge.
4. An early labour market track. Transition works better before than after damage. Therefore the state and labour market parties should start tracking where work content thins out, not just where jobs formally disappear.
5. A foundation for authenticity and verification. When synthetic content becomes cheaper to produce, the need increases for robust processes for provenance, signing and control in both the public sector and public debate.
The window is still open
The positive thing is that Sweden still has strong preconditions: relatively high digital maturity, high trust, functioning institutions and good opportunities for coordination if the will is there. But the window is not open indefinitely.
What policy misses is therefore not just that AI is important. What policy misses is that a question can be so broad and so fast that it requires different ways of working than those that normally suffice.
Source notes
The sources below are used as the basis for the essay's policy and labour market reasoning.
- AI Commission: The AI Commission's Roadmap for Sweden.
- Government: Sweden's AI Strategy, strategy document and action plan. See also Stanford AI Index 2025 and WEF Future of Jobs 2025 for broader capacity and labour market context.