Agentic AI in government faces tough question: What should machines be allowed to decide?

The United Arab Emirates has been an early adopter of artificial intelligence for nine years. It launched a national AI strategy in October 2017 and, shortly thereafter, established a ministerial role to oversee it, appointing Omar Sultan Al Olama as the world’s first Minister of State for Artificial Intelligence at age 27. Since then, the country has developed digital identity, sovereign cloud, and data-sharing infrastructure that most governments are still in the process of acquiring. The real test of agentic AI in government is whether this head start translates into tangible results.

This testing phase began this month, with over 100 federal officials attending a workshop to kick off the strategic track of the UAE’s agentic AI project. This national initiative aims to transition half of all federal government operations, services, and tasks to agentic AI models within two years. Organized by the National Committee for the Agentic AI Project, the session established implementation priorities and timelines, while initiating the development of frameworks to determine which government tasks can be delegated to AI agents.

This classification process is critical. The UAE possesses the necessary compute power and infrastructure, including an AI-driven proactive performance system that monitors over 150 million data points monthly. However, no government has yet published a defensible method for distinguishing between tasks an autonomous system can execute independently and those it should only recommend.

Huda Al Hashimi, Deputy Minister of Cabinet Affairs for Strategic Affairs, described the workshop as the operational start of deploying AI models across government activities. The program spans seven pillars: strategy and projects, foresight and strategic intelligence, policies, structures and governance, government performance, global competitiveness, and innovation in government work. Officials also discussed performance metrics and strategies to prevent duplication of efforts across entities.

Two framings of the same project

The workshop was guided by a core principle: humans lead, AI enables.

Reflecting on the project's introduction, Sheikh Mohammed bin Rashid Al Maktoum stated that the UAE would be “the first government in the world to largely deploy Agentic AI models” across its sectors and operations. He described systems capable of managing operations and executing independent actions without human intervention. While the foundational language emphasizes autonomous execution, the implementation language focuses on human primacy.

Both perspectives can coexist across a portfolio of thousands of tasks, and the gap between them is precisely what the classification frameworks aim to bridge. However, until these frameworks are published, the question of accountability for an agent’s decisions remains unresolved. The materials released so far cover pillars, training, indicators, and governance structures, but they do not address what happens when an autonomous system makes an error in a citizen’s case or where liability lies.

This is not a criticism unique to the UAE. No government has answered this question because no government has faced it before. The distinction is that the UAE will be the first to answer it at scale, under a two-year deadline.

The scale is the point

Oversight is held by Sheikh Mansour bin Zayed Al Nahyan, with a taskforce chaired by Mohammad Al Gergawi, Minister of Cabinet Affairs. In May, the Cabinet approved the first package of government services to utilize agentic AI tools, alongside the largest training program in the history of the UAE government, covering 80,000 federal employees from ministers to new hires. A national AI healthcare policy was also passed during the same session, along with frameworks making digital records the official source of core government data and requiring information to be collected once and shared securely across entities.

The pace since has been deliberate. A training event in Dubai in June attracted over 300 participants from 50 federal institutions, with the goal of identifying suitable services and procedures for agentic AI within 90 days. A separate workshop that month, organized by the Presidential Court and the Ministry of Cabinet Affairs, trained 600 staff members on the same material.

Why agentic AI in government travels beyond the Gulf

Governments across Asia and Europe are approaching the same decision from a starting point, most having an AI strategy and a national body but no specific targets. The UAE has attached specific numbers and dates, meaning that within two years, there will be public evidence regarding the actual impact of agentic AI on service delivery, headcount, and error rates.

No one else is required to run this experiment now. However, they must decide how to use the results and whether they are willing to move forward before those results are available.

The task classification frameworks are the documents to watch. When published, they will represent the first serious attempt by any state to define which decisions a government is willing to delegate to machines. Everything else in this program involves procurement and training, which are standard procedures.

See also: OpenAI president urges enterprises to hasten AI security defences

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.