The UAE wants half of its government services running on artificial intelligence agents within two years. Sheikh Mohammed bin Rashid Al Maktoum — the country's Vice-President, Prime Minister and Ruler of Dubai — first set that target in April, then returned to it at a Cabinet meeting in May with a training plan for 80,000 federal employees.
Sheikh Mohammed says the goal would make the UAE the first government in the world to run services at that scale through what is called agentic AI.
The phrase needs unpacking. Most people have met AI as a chatbot: you ask a question, it answers. An AI agent goes further. It can carry out a task end to end — pull a record, check it against a rule, fill in a form, and flag what it cannot resolve — with a person supervising rather than doing every step. The plan is to put that kind of system behind everyday government work.
The target is specific. Not half of one ministry, but half of the federal government's services, sectors and operations, measured across the board within two years.
The training is the bigger commitment. The National reported that the programme, branded UAE Government 4.0, will train 80,000 workers from ministers to junior employees. It is being built with universities and global technology firms and splits into five tracks: leadership, technical, specialist, general workforce, and trainer support. Staff will learn through a digital platform that sets a personalised path for each worker, alongside a governance framework that spells out who is responsible for what across ministries.
The rollout is meant to be phased across ministries and federal entities, with performance and impact checks deciding how widely each AI agent is adopted. That matters because the clearest early targets are likely to be repetitive, rules-heavy services where the process is well defined and the room for judgement is smallest.
For residents, the promise is plainer than the policy language: tasks that today mean a queue, a form, or a wait could be handled faster, at any hour, by a system that does not clock off. The risk is just as plain. Software making decisions about taxes, fines, or approvals raises hard questions about errors, appeals, and who answers when an agent gets it wrong. That is part of why the governance framework and a human-in-the-loop design sit at the centre of the plan.
The UAE has spent two decades moving government services online, from eGovernment and mobile government to UAE Pass and proactive service bundles. This plan is the next, sharper test: whether a government can move from talking about AI to letting supervised agents run a meaningful share of the work.
Two years is a short runway for a change this size. The targets are public, the deadline is fixed, and the training has a number attached to it.
💬 What would you want an AI agent to handle first — and what would you still want a human to sign off on?