Agentic AI in government just hit the hard part: deciding what a machine may decide
ainews ·
The UAE, an early AI adopter with a national strategy dating back to 2017, has launched a national programme to shift half of federal government operations, services, and tasks to agentic AI models within two years. More than 100 federal officials attended a workshop to set priorities and begin building frameworks that classify which government tasks can be handed to autonomous agents at all. The central challenge is drawing a defensible line between tasks an AI may complete and tasks it may only recommend, something no government has yet published. The project operates under the stated principle that humans lead and AI enables, though the founding language emphasized autonomous execution, leaving questions about accountability unresolved. Oversight includes a taskforce, and the Cabinet has approved initial services plus a training programme covering 80,000 federal employees. Because the UAE has attached a number and a date to this effort, it will be the first to provide public evidence about agentic AI's impact on government service delivery, headcount, and error rates. The classification frameworks are considered the key document to watch, as they represent the first serious attempt by any state to codify which decisions a machine may make on its behalf.
ainews