Many public service delivery organisations have started with agentic AI, and after the first redesign they run into the same problem: a process executed by agents still sits inside an organisation built for people. Departments, batches, handovers and queues exist because no one can oversee everything. Put an agent in that environment and you get a fast executor in a slow organisation. And an error in execution is not a pilot error. It lands with a citizen who is waiting for a benefit, an allowance or a permit.
The real challenge, then, is not the agent but the organisation around it. How do you define roles, rules and accountability so that employees and agents work from the same frameworks? Who decides, per task, how much autonomy an agent gets, and on what evidence? And how do you make sure citizens, and vulnerable citizens above all, experience more of the human touch in service delivery?
In this whitepaper, the follow-up to Agentic AI for public service delivery organisations, Maurice Boon, Managing Partner at Eraneos Netherlands, introduces the orgentic delivery organisation. He shows how the operating model is defined, how employees and agents divide the work side by side, and how to build safety and fallback into the architecture rather than into policy. He closes with six decisions that executives must take themselves, because a transformation of this impact cannot be delegated downwards.