Article content
For more than twenty years, Estonia’s digital identity system has been the quiet workhorse of our digital state. eID made secure online services possible and enabled the kind of administrative redesign that turned Estonia into a global model. It opened the door for things like remote signatures, cross‑border entrepreneurship, and services that work.
In recent years, Estonia has taken meaningful steps toward the next evolution of digital government through initiatives like Bürokratt, which has laid the groundwork for conversational access to public services and the eventual coordination of tasks across agencies. The ambition behind Bürokratt is real, and it reflects a state that understands the direction of technology and is actively experimenting.
Yet something fundamental is changing. We are entering a world where not only people but also AI agents act inside public systems. These agents already draft documents, retrieve registry data, compose workflows, and make decisions that resemble administrative work. They do not just advise. They act. And although Estonia is not behind in this story, the global context is moving rapidly. Ukraine’s Diia.AI has already begun delivering real services through an AI agent, and Singapore is piloting autonomous agents in licensing and social care. The UAE intends to run a fully AI-powered government in the next few years. The shift to agentic administration is no longer theoretical. It is happening.
This raises a question that Estonia cannot avoid. We built our digital state on the assumption that humans are the only actors with identity, agency, and responsibility. Everything in our infrastructure reflects that assumption, including the secure digital identity system, the Digital Signatures Act, and the core logic of administrative procedure. AI agents do not exist inside these frameworks. They cannot authenticate, sign, take responsibility, or explain their reasoning in ways that match legal standards. In practice, this means they cannot perform tasks that would meaningfully change productivity or service delivery.
This misalignment is not a technical failure. It is simply the result of a digital state designed during a different technological moment. eID was never meant to identify non‑human actors or mediate relationships between a person and the autonomous systems acting on their behalf. Yet that is precisely the relationship we now need to define. Eesti.ai outlines ambitious economic goals that depend on automating high‑value processes across the state and the private sector. But legally meaningful automation is impossible without legally recognised actors. As long as AI agents cannot be attributed, delegated authority, or held accountable, they remain tools rather than participants in the administrative ecosystem. And tools cannot drive a 25 per cent increase in national productivity.
The paradox is striking. Estonia has the world’s most interoperable digital infrastructure and one of the most trusted digital identity systems. Our culture of experimentation is strong, illustrated by the evolution of Bürokratt and the early adoption of proactive services. Yet these strengths will not carry us into the next generation of digital government unless identity itself evolves. To borrow from the logic of e‑Residency, which created a cross‑border economic community around digital identity, we now need to create an identity for the agents that will work alongside people. An AI Residency model or an Agent Identity layer would enable autonomous agents to authenticate, sign, act, and be supervised in transparent, auditable ways. It would anchor their actions in law, link them to the human or organisation they represent, and define the boundaries of their autonomy.
Estonia has done this before. We created the legal and technical foundation that enabled non‑resident entrepreneurs to operate within our digital state. That decision reshaped our economic landscape. We are now facing a moment of similar scale. If Estonia defines how autonomous agents identify themselves, how they act on behalf of users, and how they can be trusted inside critical systems, we can shape the global standards for the machine economy just as we once shaped the standards for digital governance.
The rest of the world is not waiting—some move because of necessity, others because of strategic ambition. Estonia moves best when it moves early and with clarity. Our next leap will not come from more chatbots or more data connections. It will come from recognising that identity, the very heart of our digital state, must now extend beyond humans. Without this, the potential of AI in public administration will remain out of reach. With it, Estonia can once again lead the world into a new era of digital governance.
