● Model comparison
Agentic vs cognitive government
Cognitive government uses AI as an analytical core: it understands and predicts. The Agentic State makes AI a protagonist that carries a situation to a result, and adds what cognitive government lacks: a citizen agent and built-in contestability of decisions.
Intelligence without agency and rights is surveillance, not service.
Cognitive government: smart but observing
The cognitive model puts analytics at the centre: models predict demand for services, risks, behaviour. The state “thinks” about the citizen better and earlier. But decision and execution in this model often sit outside explicit rights and contestation procedures.
The danger is not the system’s stupidity but its opaque insight: the more accurate the forecast, the stronger the temptation to act pre-emptively without explanation or consent.
The Agentic State: acting and accountable
The agentic model does not reject analytics — it embeds it in action and in a frame. AI becomes an agent that carries the situation, but every material action must be explainable, recorded in an audit trail and open to contestation.
The decisive addition is the citizen agent. In the cognitive model the person is an object of analysis; in the agentic model they have their own representative who demands explanations and contests decisions.
Where the line runs
Cognitive government answers “what the system knows”. The Agentic State answers “what the system does”, “by what right” and “how to contest it”. This is a shift from epistemology to the constitutional mechanics of AI in power.