The UAE is running an experiment no other government has been willing to run: it has committed to shifting half of all federal government operations, services, and tasks to agentic AI within two years. The clock started this month, when more than 100 federal officials gathered for a workshop to launch the strategic track of the national programme. Huda Al Hashimi, Deputy Minister of Cabinet Affairs for Strategic Affairs, called it the operational beginning of deploying AI models across government activity.
What the UAE already has is enviable. Nine years of groundwork—a national AI strategy from October 2017, the world’s first minister of state for AI in Omar Sultan Al Olama, a sovereign cloud, digital identity, and an AI-powered performance system tracking over 150 million data points monthly. What it doesn’t have, and what no government has, is a published method for deciding which tasks an autonomous system may complete versus which it may only recommend.
That classification exercise is where this gets interesting. Sheikh Mohammed bin Rashid’s founding language described systems that run independent actions without human intervention. The implementation language says ‘human leads, AI enables.’ Both can be true across thousands of tasks, but the gap between them is precisely what the frameworks must close. The material released so far covers pillars, training, indicators, and governance structures. It says nothing about what a citizen does when an autonomous system gets their case wrong.
The scale is deliberate. A taskforce chaired by Mohammad Al Gergawi oversees the effort. In May, the Cabinet approved the first package of services, alongside a training programme covering 80,000 federal employees—the largest in UAE government history. A June Dubai training event drew 300 participants from 50 federal institutions, with a 90-day goal to identify suitable services. Another workshop put 600 staff through the same material. Governments across Asia and Europe are watching from a standing start, most with strategies and no numbers attached. The UAE has attached a number and a date, which means within two years there will be public evidence about what agentic AI actually does to service delivery, headcount, and error rates. The classification frameworks, when published, will be the first serious attempt by any state to write down which decisions it will let a machine make on its behalf.
