Eesti.ai – the next stage in Estonia’s digital transformation

Estonia launched Eesti.ai in 2026 – a programme in which the adoption of artificial intelligence is treated as an instrument of economic policy. The declared goal is ambitious: to double labour productivity by 2035 and increase GDP by 50% within ten years. The approved pool of fifteen flagship projects operationalises this ambition, while at the same time opening up a number of questions regarding the feasibility of the expected effects, their distribution within society, and the legitimacy of the decision-making process. The outcome of this public policy experiment will be of significance for the whole of Central Europe, including Poland, where many questions about the shape of AI policy remain open.

The origins and logic of Eesti.ai. The Estonian government launched the Eesti.ai programme at the beginning of 2026, with implementation entrusted to the Government Office (Riigikantselei), in line with the Estonian practice of placing key strategic initiatives under the Prime Minister’s Office. According to the rationale, the programme responds to the problem of a shrinking labour force in an ageing population. The logic of the response, as Prime Minister Kristen Michal puts it, is as follows: since the labour force is not growing, the economy can only grow if the value of each employee’s work increases – and this value is to be raised precisely by artificial intelligence, which makes it possible to perform tasks with higher added value. Eesti.ai is the next stage in the digital state that Estonia has been consistently building, now extended to include the systemic adoption of artificial intelligence. For several years now, Estonia has been implementing successive AI action plans, the so-called Kratt plans, the latest of which – Kratt 2024–2026, led by the Ministry of Economic Affairs and Communications – has a budget of 85 million EUR. However, an important element is the clearly defined macroeconomic objective: the Eesti.ai programme explicitly links digital policy with measurable targets: doubling labour productivity by 2035, increasing GDP by 25% within five years and by 50% within ten years, which would mean an additional 20 billion EUR in GDP.

Operationally, the programme is led by Kirke Maar, formerly the head of AI & Robotics Estonia (AIRE) – one of the European Digital Innovation Hubs (EDIHs), which serve as support points for EU enterprises and the public sector in digital transformation, in this case, in the field of artificial intelligence and robotics. The advisory body of Eesti.ai is a council that functions as a government committee: the Prime Minister participates in its meetings, and sessions are held quarterly. At the time the programme was established, the council consisted of six people – three technology entrepreneurs: its chair, Markus Villig of Bolt, Taavi Madiberk of Skeleton Technologies, and Kaspar Korjus of Pactum; two investors: Sten Tamkivi, partner at the Plural fund, and Jaan Tallinn, co-founder of Skype; as well as Linnar Viik, one of the architects of Estonia’s digitalisation. However, the original composition of the council met with public criticism. Karen K. Burns – co-founder of the Estonian AI company FYMA and a lecturer at Cambridge Judge Business School – pointed, among other things, to the lack of specialists in ethics, law, and AI research, as well as the absence of women. In response, programme head Kirke Maar announced that the council would be expanded – and at the next meeting, three new members joined: Anima Anandkumar of Caltech, former director of machine learning research at NVIDIA; Risto Uuk, Head of EU Policy and Research at the Future of Life Institute; and Sirli Männiksaar, Chair of the Management Board of Ericsson Estonia.

The mechanism and the first pool of projects. In April 2026, the council approved fifteen flagship projects in six areas: digital skills, enterprises, the public sector, education, health, and computing infrastructure. In the area of skills, an initiative to create the “most AI-aware nation” was launched, including training for 100,000 people over eighteen months. Enterprises are given access to three support instruments – from mentoring programmes, through AI implementation vouchers worth 20,000 EUR, to development grants of up to 500,000 EUR. In the public sector, artificial intelligence serves to support the legislative process, public procurement, and administration, while a separate project maps the demand for national computing infrastructure. In education, the programme assumes the expansion of AI Leap to primary and vocational schools and the preparation of an Estonian dataset for large language models (“IEŚ Commentaries”, No. 1509). In healthcare, projects include, among other things, the integration of AI with the family doctors’ platform and the automatic transcription of documentation in ambulances.

Analysing the selection of projects reveals two features of the programme: first, an operational horizon oriented towards measurable effects within six to twelve months. Second, a logic of concentration, which the programme’s authors themselves define as the selection of a small number of highly effective projects. The first pool includes the fifteen aforementioned projects, with clearly assigned ministerial responsibility – which, compared with previous Kratt plans that financed dozens of parallel pilot applications, represents a significant shift in the proportions.

The full funding pool has not yet been announced. The government has stated that this will be presented in spring or summer 2026, while in the first phase, the programme is to use existing support instruments, including funds provided under the Kratt 2024–2026 plan and EU programmes. Estonia’s public finances, however, are under pressure: the Estonian central bank forecasts Estonia’s general government deficit in 2026 at 4.4% of GDP – the highest level in three decades, excluding the pandemic period – which, alongside rising defence expenditure of more than 5% of GDP, limits the budgetary space for new commitments.

The questions raised by the programme. The scale of Eesti.ai’s ambition has triggered numerous questions in Estonia’s public debate. The first concerns the realism of the assumed productivity growth. The declared macroeconomic effect is based on the assumption that AI adoption translates directly into an increase in labour productivity; however, the latest empirical studies remain cautious on this matter. Moreover, Estonia is already the regional leader in AI adoption among enterprises – 23.4% of companies were implementing this technology in 2025. Growth from a higher starting point usually requires greater expenditure per unit of increase than in countries starting from a lower base, and the returns from further implementations may be lower.

Another question concerns the redistribution of effects. Council member Sten Tamkivi himself, in parallel with the launch of the programme, pointed to two risk scenarios whose occurrence would change the assessment of the entire initiative. The first is the “super-productivity paradox”: the share of the information technology sector in GDP grows, but employment in it falls as a result of automation, which, in the Estonian tax model, based on labour taxation, means an erosion of the fiscal base. The second is the “10X elite paradox”: a narrow group of AI specialists accounts for a significant part of the state budget, making the system extremely dependent on their decisions. Nevertheless, these are questions about the tax order, income distribution, and the social contract – areas that go beyond the current scope of the Eesti.ai council’s activity.

Conclusions. Eesti.ai stands out from other national AI strategies in the European Union in three respects: its explicit linking of AI policy with a response to demographic change, the concentration of resources on a small number of projects with measurable impact, and its institutional embedding at the level of the Government Office, with clear involvement from the technology sector. From the perspective of Central European countries, three observations can be drawn from the programme.

• First, the Eesti.ai model is not transferable to Central Europe in any simple sense. Estonia is launching the programme from a level of 23.4% AI adoption among enterprises in 2025 – compared with 8.4% in Poland, 8.6% in Bulgaria, and 5.2% in Romania. This difference in the starting position should be kept in mind when formulating AI policy in the other countries of the region.
• Second, the model of a “government acting like a start-up” – short iterations, measurable effects, and close cooperation with the private sector – has operational advantages but also reveals limitations: the underrepresentation of voices from outside the technology ecosystem and questions about the representativeness of the decision-making process. The Estonian debate around the composition of the council shows that, even in a country with an established digital culture, these issues require separate reflection.
• Third, the next eighteen months constitute a test of the programme’s internal logic. Estonia is undertaking an empirical verification of a hypothesis whose results remain disputed among economists: whether the broad adoption of AI actually translates into aggregate productivity growth across the entire economy. The outcome of this test – regardless of its direction – will be significant both for the countries of Central Europe and, more broadly, for the entire European Union.