I joined the federal civil service two months after the June 2013 protests, which demanded better public services. The question that has stayed with me since is simple to ask and hard to answer: why does the citizen remain dissatisfied, even after everything that has been done?
What was done was not small. Brazil went from fewer than a thousand federal digital services to more than four and a half thousand in a few years, reached a leading world position in digital government maturity, and generated savings in the billions for citizens and for the state itself. These are real advances, built by thousands of civil servants who devoted their careers to making the state more accessible.
Even so, trust was not won back. That is the starting point of the book I am finishing, Agentic Digital Government. And the answer I found changes the way we look at public sector digital transformation.
The citizen does not compare government with yesterday's government
The citizen compares government with the best digital experience they had the day before. When someone solves a problem on their phone on a Sunday night and has the solution by Monday morning, it recalibrates their whole expectation of what good service means, including from the state.
Digitisation narrowed the gap between government and citizen. But the market raised expectations faster than digitisation could keep up. And now, with artificial intelligence agents that complete complex tasks in seconds, that expectation has accelerated again.
User satisfaction is a moving target. And it is moving away.
When citizens themselves are asked what those in government must do to legitimise their authority, the answer fits in two words: understand what people need, and have a real impact on their lives. Empathy and impact. Neither of them is, in itself, a technology question.
The map is still right. It has simply become incomplete
The international digital government frameworks that guided dozens of countries, Brazil among them, remain valid. The state needed to move from a counter posture to a platform posture, and it did. The problem is that those frameworks were designed for an expectation the citizen has already moved beyond.
Expectation has shifted three times. First the citizen wanted access. Then convenience. Now they want agency: someone to solve it for them, under their command. The book proposes the missing layer that answers this third expectation, and gives it the name Agentic Digital Government.
It does not replace what exists. It is a layer laid over it, in the same way digital was laid over electronic. The state that results stops merely responding well and starts acting as the citizen's copilot.
No, it is not just building a chatbot
That is the most common misconception I come across. Placing a conversational assistant in front of an old service produces a new interface over the same process as ever. The citizen still has to integrate, alone, a state that ought to work as one.
A single agent, however sophisticated, cannot be the government.
What the book describes is something else: a form of mediation between state and citizen that speaks the language of whoever is asking, understands the context of whoever arrives, handles several needs in the same interaction and lets the citizen know, before formalising anything, whether what they hold meets what the state requires. Behind it lie orchestration, governed data and civil servants in a different role, not a tool.
And there is a consequence that tends to surprise people expecting the opposite. When work that does not require human presence is absorbed by agents, the civil servant is freed for the atypical case, the decision under ambiguity, and the support of those in vulnerable situations.
AI makes government more human, not less.
What a leader would want at the end of the road
I wrote the book with one person in mind: the public leader who has already been through digitisation, already invested, already delivered, and still cannot show that the citizen's life has changed. For that person, the book describes the destination. The state they would want to have built when they look back.
But the destination is not reached through technology. The final chapters deal with what sustains the shift: maturity in citizen experience, which delivers empathy, and a management model able to turn understanding into real impact, at the time of need and at the scale society demands. I call that model Exponential Public Management.
The traditional manager asks how to deliver what was planned. The exponential manager asks how to solve the public problem at scale.
That kind of leader exists, but is rare. Not for lack of talent in the Brazilian public sector, but because training institutions still mostly prepare a different profile. The exponential manager is still formed by exception, with few spaces dedicated to them.
NEXT C-LEVEL exists to open that space. The book describes where to arrive. The programme trains the people who will lead the journey.
A note on timing
Agentic AI is still a frontier, not an established reality in the public sector. Part of what the book proposes will be validated quickly. Another part will need adjusting once governments start implementing at scale.
What does not change is the starting point. Citizens want to be understood and want a positive impact on their lives. That demand existed before artificial intelligence and will continue after it. Agentic AI is the most powerful instrument ever available to meet it.
Using it well is a management choice, not a technology choice.