The agentic state: why Europe must act now

This article contributes to the Friends of Europe’s Peace, Security and Defence programme and Ukraine Initiative. Its insights feed into the Jacques Delors Friends of Europe Foundation’s broader Spending Better initiative, which seeks to optimise defence spending, strengthen European institutions and capabilities, and enhance Europe-wide preparedness and societal resilience.

Countries that built the most advanced digital governments may now be the least prepared for what comes next. This paradox is becoming clear as governments confront the shift to agentic AI. These are states that spent two decades building sophisticated digital systems – but now may need to fundamentally rethink how they operate.

Strong digital states create systems that work: trained civil servants, adapted citizen expectations and political credibility. Yet in a moment of deep technological change, these strengths can become constraints. The habits and structures of digital government were built for a different model, making the transition harder for leaders than for those still building from scratch.

I saw this firsthand during my six years as Deputy Minister for Digital Transformation in Ukraine. We were constantly rethinking how the government should work, starting from real user needs rather than legacy systems. That habit of continuous reinvention is exactly what the shift to agentic AI now demands. More mature governments, by contrast, often have to unlearn old ways before they can move forward.

Agentic AI is not just another tool. It challenges how governments operate. Previous technologies changed access to services; agentic AI changes how those services are actually delivered. It can plan, act and coordinate, forcing governments to question whether existing processes should exist at all. Most importantly, it addresses one of the state’s biggest constraints: talent. Agentic systems expand the ability to manage complexity and decision-making at a scale no workforce can match.

This shift is already visible. In procurement, agentic systems can find suppliers, manage compliance and detect risks at scale. In anti-corruption, they enable continuous monitoring instead of periodic audits. For citizens, agents can navigate complex systems on their behalf, removing the need to understand bureaucratic structures.

Ukraine’s experience in this transition is different – not just because of the war, but because it developed the ability to continuously adapt under pressure. During the 2022 full-scale invasion, resilience came from the ability to move fast, relocate critical systems into the cloud within days, reconfigure institutions while under attack, and sustain state functions in conditions no continuity plan had anticipated. This is sovereignty in its most operational form – and it is directly relevant to how governments should approach AI: not through rigid control, but through the capacity to adapt quickly as conditions evolve.

For European governments, this transition requires both central capacity and distributed capability. A strong core is needed to build shared platforms, governance frameworks and standards that can scale across the state. At the same time, institutions must gain hands-on experience with these tools to become informed and demanding users.

Governments also need to clearly define what they build themselves, what they open to the market, and what they regulate. Failing to set these boundaries risks leaving them to external platforms whose incentives do not fully align with the public interest.

Those who treat agentic AI as just another technology upgrade risk falling behind. The consequence is a slow erosion of trust as the gap grows between what citizens experience in the private sector and what they expect from the state.

Digital leaders have built something real. The question now is whether they can use it as a foundation to move forward, or whether it becomes a constraint.

This article draws on insights from “FORWARD with Valeriya Ionan”, a podcast for decision-makers and builders. The full conversation with Luukas Ilves, former Chief Information Officer of Estonia, that informed this article is available here. The views expressed in this #CriticalThinking article reflect those of the author(s) and not of Friends of Europe.