“Why talk about it now? Because the political stakes have changed. Citizens judge the State by their last interaction with it, and every failed interaction feeds the narrative that government is structurally useless.”

“We will build deliverables with teeth.”

Government Tomorrow Forum: How do you see the “hey government” concept and why do you think it is important to talk about it today?

Alexandre Pointier: The relationship between citizens and the State has always run in one direction: the citizen does the work. You learn the administration’s language, find the right office, fill in the forms, and carry the burden of proof. These administrative burdens fall hardest on the people who need the State most. In France, roughly one in four people entitled to certain social benefits never claim them.

“Hey government” names the inversion of that relationship. For the first time, the technology exists for the State to come to the citizen. What it means is that the State can anticipate a life event, trigger a benefit automatically, answer in plain language, act rather than instruct. The OECD’s Digital Government Outlook 2026 is clear on the paradox: most governments now recognise proactive service delivery as a goal, yet proactiveness remains among the least implemented dimensions of digital government.

Why talk about it now? Because the political stakes have changed. Citizens judge the State by their last interaction with it, and every failed interaction feeds the narrative that government is structurally useless. That is the narrative that produced the chainsaw politics embodied by DOGE in the United States. I just published a book arguing that the crisis of the State is not a resource problem but a capacity-to-act problem: the State decides without managing to act. Agentic public services are where that abstract question becomes concrete for every citizen. Democracies probably won’t survive if they completely miss that step.

GTF: What are the less obvious sides of the agentic AI-powered public services that governments might easily oversee?

AP: The first trap is the demo illusion. It is a trap we observed with many digital transformations over the last decades. You change the front end but maintain a broken back office. The hard part is never the agent but the plumbing behind it: data interoperability, legal framework for data sharing, etc.

The second is accountability. When an agent acts on your behalf and gets it wrong, who is accountable, and what recourse does the citizen have?

The third is the people inside the state. Agentic services will reshape the role of civil servants. If you don’t redesign jobs, training, and accountability around that new role, you get either resistance or rubber-stamping.

The final trap concerns those who will never say "Hey Government". Human-centricity is measured at the margins: people without smartphones, people with problems that don’t have a standardized answer, etc… Keeping human channels open is critical.

GTF: Should this type of public service be more government-driven, or private sector-driven and why?

AP: Posed as either/or, it’s the wrong question. The right question is: who owns which layer.

The State will ultimately be responsible for the service provided, the protection of citizens’ data and the continuity of service. So it must own the digital identity, the data infrastructure and the calibration of the model.

On top of those public rails, the private sector should drive the innovation: interfaces, models, user experience, speed.

But three conditions are non-negotiable. Open standards and reversibility, so no vendor becomes a chokepoint the State cannot exit. Enough in-house engineering capacity for the State to be a competent buyer. And public accountability for outcomes, whoever operates the software.

GTF: What do you expect from the workshop conversation?

AP: Three things. First, the confrontation of radically different starting points: the UAE building at greenfield speed, Ukraine running Diia under wartime conditions, Germany and France carrying decades of administrative legacy… A framework that survives contact with all three contexts is worth publishing.

Second, honesty. Two hours, a small room, Chatham House rules. Anonymized honesty. The format is designed for people to say what actually failed and what actually worked. I expect disagreement, particularly on where the public-private line sits. That disagreement, written down clearly, is the most useful thing we can hand to every government that will face these choices in the coming years.

Third, a deliverable with teeth. The agentic-readiness checklist should measure the back office, not the front end. The questions a minister or CIO can ask on Monday morning: can the data legally flow between agencies? Who is accountable? Can citizens appeal? Can you exit your vendor? Where to start before having the full-blown “Hey Government” agent? A checklist like that tells serious companies what serious governments will demand.