UAE targets AI agents delivering half of federal services within 3 years

Key Points

1. AI agents will deliver 50% of UAE government services within three years, the AI minister said.
2. A federal control tower will log and audit every action and decision taken by those agents.
3. A risk review found 70% of services suitable; higher-risk ones stay off limits for now.

The latest:

Half of the UAE’s government services will be delivered by artificial intelligence agents within three years, Omar Sultan Al Olama, Minister of State for Artificial Intelligence, Digital Economy and Remote Work Applications, said at a panel on the future of AI and the limits of trust in autonomous agents. He also disclosed a federal oversight system, the AI Agents Control Tower, built to keep the technology inside supervised boundaries.

Details:

• The target: Al Olama said the UAE is moving toward a qualitative shift in its government services system, with AI agents handling 50% of services over the next three years. He framed the push as part of the leadership’s direction to accelerate adoption of advanced technologies and raise the efficiency of government work.
• The control tower: According to Al Olama, the federal government operates a centralized monitoring system for AI agents, and every agent deployed within its scope must connect to it. Each action an agent executes and each step or decision it takes is recorded and subjected to continuous auditing.
• Human oversight: The system provides continuous supervision and audit mechanisms to understand what decisions the intelligent systems reach and how they execute them, Al Olama said, alongside human oversight. He described the aim as strengthening governance and accountability and preventing the technology from drifting from its assigned objectives.
• The risk screen: Al Olama said the expansion rests on a detailed assessment of each service and its risk levels. A full inventory showed roughly 70% of services carry varying degrees of suitability for AI agents, against 30% involving higher risks that rule out expansion at this stage.
• Where it starts: The evaluation examined what could go wrong if an AI agent reached an incorrect decision, Al Olama said. Passport and document renewals carry relatively low risk, making them better suited to the technology, while others demand caution because of the nature of their decisions and their consequences for individuals.
• The red line: Al Olama said any negative impact on a person’s life or livelihood from an agent’s decision is a red line for the government, particularly at a stage when the technology’s full real capabilities remain unknown. Human life and livelihood, he added, cannot be reduced to figures or financial gains.
• The success metric: The government does not treat AI as merely a tool for raising productivity or cutting spending, Al Olama said. Quality of human life sits at the front of the criteria for measuring digital transformation, and gains achieved at a person’s expense do not count as an achievement.
• Procurement limits: Al Olama said AI in government cannot rest on buying software from companies without ensuring they understand the nature of government services and the specificity of their applications. Such systems, he said, must operate under responsible bodies capable of understanding their functions and monitoring their performance.
• Public trust: Preserving society’s trust is a core pillar of the government transformation drive, Al Olama said, and any expansion in AI use must be paired with clear oversight systems that guarantee decisions are sound and protected.

Between the lines:

The 70/30 split is the operative constraint on the 50% target: the government has already mapped which services it will not automate, meaning the three-year goal sits well inside the pool it judged safe. Pairing the target with a mandatory logging system suggests auditability, not speed alone, is the condition for scaling.

What’s next

Watch which federal entities connect their agents to the control tower first, whether document-renewal services lead the rollout, and whether the 30% higher-risk category is reclassified as the technology’s capabilities become clearer.