UAE Begins Historic Push to Move 50% of Federal Government Work to Agentic AI
The United Arab Emirates is moving beyond AI assistants and chatbots.
It now wants artificial intelligence systems capable of planning, executing tasks, coordinating actions, and in some cases making decisions with greater autonomy to become part of the machinery of government itself.
More than 100 federal officials joined a workshop in Dubai in August to launch the strategic track of the UAE's Agentic AI Project, a national initiative targeting the transformation of 50 percent of federal government operations, services, and tasks into agentic AI-driven models within two years.
The workshop, organised by the National Committee for the Agentic AI Project, focused on something much more consequential than choosing an AI platform.
Officials began developing frameworks for determining which government tasks should be handled by AI, how those tasks should be classified, and how federal agencies should coordinate their deployments without duplicating systems.
That may become the most important part of the entire experiment.
The technology can be purchased.
The harder question is:
What should a government actually allow an AI agent to do on its behalf?
From AI Strategy to Agentic Government
The UAE has spent nearly a decade preparing for this transition.
In October 2017, it launched the UAE Strategy for Artificial Intelligence, positioning AI as a central component of its long-term government and economic strategy.
That same period saw the appointment of Omar Sultan Al Olama as Minister of State for Artificial Intelligence, making the UAE the first country to establish such a ministerial position. His portfolio was later expanded to include the digital economy and remote work applications.
But the government's latest initiative represents a significant escalation.
On April 23, 2026, Sheikh Mohammed bin Rashid Al Maktoum, Vice President and Prime Minister of the UAE and Ruler of Dubai, announced a government framework designed to shift 50 percent of government sectors, services, and operations toward Agentic AI within two years.
Unlike conventional generative AI, which typically waits for a user to ask a question or issue a prompt, agentic AI systems can potentially:
- interpret an objective
- break it into multiple tasks
- retrieve information
- interact with software systems
- initiate workflows
- evaluate results
- take additional actions
- escalate exceptions to humans
In other words, the ambition is not simply for AI to answer government employees.
It is for AI increasingly to perform government work.
"Human Leads, AI Enables"
The August workshop introduced an important principle for the programme:
"Human leads, AI enables."
That statement matters because it establishes, at least at the policy level, that artificial intelligence is intended to operate under human-led governance.
But it also creates one of the central questions surrounding the project.
When the initiative was originally announced, Sheikh Mohammed described Agentic AI systems capable of managing operations and carrying out sequences of actions without human intervention. The UAE government also described the programme as enabling more autonomous execution and decision-making.
Those two concepts are not necessarily contradictory.
An AI agent could autonomously execute routine administrative processes while humans retain authority over high-impact decisions.
The challenge is deciding where exactly that boundary sits.
That is why the task-classification framework being developed by the UAE could ultimately be more important than the AI models themselves.
Seven Areas of Government Are Being Examined
The strategic government track covers seven major areas:
- Strategy and projects
- Foresight and strategic intelligence
- Policies
- Structures and governance
- Government performance
- Global competitiveness
- Innovation in government work
Officials at the August workshop also discussed implementation timelines, performance indicators, task classification, coordination between agencies, and methods for reducing duplication.
Huda Al Hashimi, Deputy Minister of Cabinet Affairs for Strategic Affairs, described the programme as entering its operational stage for applying AI models to strategic government work.
That makes this different from many government AI initiatives.
The UAE is not limiting agentic systems to answering citizen inquiries or automating simple transactions.
It is exploring AI in areas involving policy development, strategic foresight, institutional performance, and government decision support.
UAE Already Has AI Agents Working in Government
The project is not purely theoretical.
During a national government retreat in May, the UAE unveiled its first group of specialised government AI agents.
They included:
- Procurement AI Agent
- Tax Auditing AI Agent
- Customer Happiness AI Agent
- Technical Support AI Agent
The procurement agent is designed to assist sourcing and procurement workflows, while the tax auditing agent supports tax reviews and data verification.
These are significant testing grounds because procurement and taxation involve structured rules, large volumes of data, financial consequences, and established audit processes.
They also provide an early indication of how the UAE may divide responsibilities between humans and autonomous systems.
80,000 Government Workers Will Be Trained
The transition is also a massive workforce programme.
In May, the UAE Cabinet approved an Agentic AI skills programme covering approximately 80,000 federal government employees.
Training will span five categories:
- leadership
- technical employees
- specialists
- general workforce
- train-the-trainer participants
The government also plans to develop an AI-powered digital training platform that can assess employees' capabilities and provide personalised learning programmes based on their jobs and skills.
This is important because the UAE's strategy does not publicly frame Agentic AI simply as a replacement for government employees.
Instead, government workers are expected increasingly to become users, supervisors, designers, evaluators, and managers of AI-powered systems.
The Data Infrastructure Is Already Enormous
The UAE is also entering the agentic era with an established government data and performance infrastructure.
In July 2025, the government launched its Proactive Government Performance System, an AI-powered platform capable of processing more than 150 million data points every month.
The system is designed to generate more than 50,000 proactive insights annually and potentially save more than 250,000 working hours per year.
It uses AI analytics and predictive technologies to monitor government performance, track strategic objectives, generate reports, and help leaders anticipate future challenges.
That infrastructure helps explain why the UAE may be able to experiment with agentic government faster than countries that are still attempting to integrate databases across separate ministries.
Agentic AI becomes much more powerful when government systems can securely access consistent, interoperable data.
But Who Is Responsible When an AI Agent Gets It Wrong?
This is where the UAE experiment becomes globally important.
Imagine an AI system involved in:
Taxation.
Licensing.
Procurement.
Healthcare administration.
Immigration services.
Government benefits.
Regulatory compliance.
Policy analysis.
What happens when the AI makes an incorrect decision?
Who carries legal responsibility?
The software provider?
The government department?
The official supervising the agent?
The minister responsible for the agency?
And what mechanism allows a citizen to challenge an automated outcome?
These questions become increasingly important as AI moves from recommendation to execution.
The UAE has already adopted broader AI principles covering responsible use, privacy, data protection, security, and compliance with legislation.
But large-scale Agentic AI creates a different governance problem because an agent may potentially perform a sequence of actions rather than simply generate a single answer.
Every additional degree of autonomy raises questions about logging, explainability, auditability, access controls, escalation, human review, and accountability.
The Most Important Document May Not Be an AI Model
Much of the world's attention will naturally focus on what technology the UAE selects.
But the more important development may be the task-classification framework now being discussed.
That framework could eventually answer questions such as:
Which government tasks may AI execute automatically?
Which require human approval?
Which may only receive AI recommendations?
Which decisions should never be delegated to an autonomous system?
That is the governance architecture every country pursuing Agentic AI will eventually need.
Because there is a major difference between allowing an AI agent to schedule an appointment and allowing one to determine whether a citizen qualifies for a government benefit.
The UAE Is Giving the World a Two-Year Experiment
Most governments already have AI strategies.
Many have AI councils.
Many are experimenting with generative AI.
Far fewer have attached a measurable percentage and deadline to the transformation of government operations.
The UAE has.
50 percent.
Two years.
That means its success and failures could provide valuable evidence about whether Agentic AI actually improves government productivity, service delivery, decision-making, operating costs, and citizen experience.
It could also expose new forms of risk.
That makes the UAE one of the world's most important real-world experiments in AI-powered government.
