Governing the Cognitive State: AI in Government

October 2026

The report examines the growing use of AI in government and the governance challenges it creates. Governments are moving beyond digitization toward the Cognitive State, where AI increasingly supports public-service delivery. The report’s central message is that governments must build the capacity to govern AI at scale while preserving transparency, due process, public trust and administrative sovereignty.

Three pillars for AI in government

• Delivering public value through service transformation and automation
  AI in government should deliver measurable benefits, including faster decisions, fewer errors, better access and more consistent services. The report identifies disciplined service design—bounded use cases, clear performance indicators, reliable data and effective human oversight—as the strongest path from experimentation to scale. It also highlights the “pilot trap,” where governments continue launching pilots without converting them into sustainable public services.
• Governing AI through procurement and contracting
  The report shows that AI governance is also a contracting issue. Governments should use procurement to embed auditability, data portability, performance benchmarks, and remedies into AI systems from the outset. Trust must be written into contracts and vendor obligations.
• Ensuring operational accountability after deployment
  Many AI-related risks emerge only after systems become operational. The report therefore calls for continuous assurance through named accountable owners, ongoing monitoring, incident-escalation pathways, accessible redress and periodic reassessment. As agentic systems gain the ability to undertake multi-step actions, governments must also establish clear limits on delegated authority and identify decisions that must remain under human control.

The report also sets out a three-stage roadmap for governments at different maturity levels. The Foundational stage focuses on priority use cases, procurement guidance, and a minimum accountability model. The Accelerating stage focuses on scaling AI into priority services and introducing monitoring and public AI registers. The Leadership stage focuses on reusable government AI components, shareable procurement clauses, and mature assurance regimes with re-evaluation triggers.