Agentic states: governments begin adopting AI to deliver services

After businesses and consumers, AI agents are now making their way into the public sector. Simone Maria Parazzoli, co-founder of The Agentic State—a non-profit that studies the use of agentic AI by governments—says the world is moving toward an era of “agentic states,” a transition expected to accelerate over the next two years.

Parazzoli presented this vision on Wednesday, 30, at Atlantic Convergence in Lisbon. The organization has been tracking this topic for about a year and a half, since publishing a report that projected how governments operated by AI agents would function, based on interviews with 21 public policy makers from 15 countries.

Recently, the organization launched an online observatory to map similar initiatives worldwide. In just 48 days, the tool recorded 360 projects in 56 countries, 59% of which are already operational.

Advanced examples: Estonia, United Arab Emirates, and Ukraine

Among the most advanced cases, Parazzoli highlights three nations. Estonia is developing a digital identity system for AI agents, which will allow them to represent Estonian citizens when requesting public services.

The United Arab Emirates took a more ambitious approach: the government announced a goal of having half of public services coordinated by AI agents within two years.

Ukraine, meanwhile, has begun implementing an agentic state architecture based on three agents: one dedicated to serving the population, another to support civil servants, and a third focused on compliance.

Strategies: offensive or defensive

According to Parazzoli, governments can follow two paths. In the offensive approach, the state proactively launches public services operated by AI agents. In the defensive approach, it reacts to an existing demand: citizens using tools outside public authority control, such as ChatGPT and Claude, to navigate bureaucracy.

Of the 360 projects mapped, 73% follow the offensive strategy, 24% the defensive strategy, and 3% combine both.

The defensive movement already has a concrete reason behind it. According to Parazzoli, some countries are facing saturation of their digital public services due to the volume of access requests made by agents created by citizens using commercial AI tools.