More than 100 federal officials met in the UAE on 9 August 2026 to launch the strategic workstream of the country's national agentic AI project, the phase that moves a May policy announcement into scheduled delivery. The programme, organised by the National Committee for the Agentic AI Project, commits the federal government to converting 50 percent of its operations, services and tasks to agentic AI models within two years — placing the target date in 2028.
Huda Al Hashimi, Deputy Minister of Cabinet Affairs for Strategic Affairs, described the session as the operational start of deploying AI models to strengthen the UAE's global leadership in future-ready governance. The workstream sits under the UAE Government 4.0 framework and covers seven pillars: strategy and projects, foresight and strategic intelligence, policies, structures and governance, government performance, global competitiveness, and innovation in government work.
Key Highlights
- 50 percent of federal government operations, procedures and services are to be converted to agentic AI models within two years, under a framework revealed on 15 May 2026.
- The 9 August session launched the strategic workstream — policy formulation, foresight and decision-making — after an earlier operational phase focused on services.
- Seven governance pillars are in scope, governed by the stated principle "human leads, AI enables."
- Two central bodies have set themselves a higher bar: the Ministry of Cabinet Affairs targets 75 percent of its operations and the Presidential Court 75 percent of its services.
- Federal entities were given a 90-day window to identify and launch a service redesigned around agentic AI — a clock that started at the June workshop.
Details
The programme was revealed on 15 May 2026 by Sheikh Mohammed bin Rashid Al Maktoum, Vice President and Prime Minister, under the directives of President Sheikh Mohamed bin Zayed Al Nahyan, with oversight assigned to Vice President Sheikh Mansour bin Zayed. The National Committee for the Agentic AI Project is chaired by Mohammad bin Abdullah Al Gergawi, Minister of Cabinet Affairs, with Haitham Al Rais serving as secretary-general.
Execution began on 19 June 2026 with a Dubai workshop attended by 600 employees from the Presidential Court and the Ministry of Cabinet Affairs, following an earlier session that brought together officials from 50 federal entities. That workshop set the pattern the rest of the programme follows: each entity selects one specific service or operational process, redesigns the journey around agentic AI, and launches it within 90 days. The rollout is sequenced as assessment and readiness, capability building, pilot deployment of specialised agents, and full-scale rollout.
"Agentic AI does not replace people, it makes them more capable, and opens wider possibilities for creativity, innovation and impact," Al Gergawi said at the June session. The government has paired the mandate with a workforce programme covering tens of thousands of federal employees and a "Top 3 AI Agents Award" intended to reward measurable impact rather than pilot volume.
The 9 August workshop extended that model upward, from citizen-facing services into the institutional machinery: how policy is drafted, how strategic intelligence is produced, how performance is measured. Officials outlined timeline priorities, a framework for classifying which tasks are eligible for agent execution, and mechanisms for cross-entity integration.
Impact
For businesses selling into the Gulf, the significant word in this announcement is not "AI" but "procurement." A stated conversion percentage with a date attached becomes a budget line, and a 90-day launch requirement per entity becomes a stream of scoped integration work. Fifty federal entities each shipping at least one redesigned service creates demand for exactly the unglamorous layer that agent deployments actually fail on: connecting agents to the systems of record, government platforms and legacy databases where the data lives.
The regional context supports the ambition. Confluent's 2026 Data Streaming Report found 38 percent of organisations in the UAE and Saudi Arabia already running agentic AI in production, the highest rate of any market surveyed and six points above the global average — but the same survey found two-thirds naming data infrastructure, not model capability, as the barrier to scaling. We covered that gap in UAE and Saudi Arabia lead the world in agentic AI in production.
The cross-entity integration mechanisms mentioned in the August session are the tell. Agents that work in isolation inside one ministry produce demos; agents that read from and write to other entities' systems produce the efficiency numbers the mandate is measured on. That is an integration problem before it is an AI problem.
Background
The UAE has been running the shortest distance between announcement and deployment in the region. Dubai Municipality said on 9 August 2026 that it is building an AI agent to issue villa building permits automatically, reading submitted drawings against the Dubai Building Code and compressing an approval from days to minutes — see Dubai Municipality builds an AI agent that issues villa building permits in minutes. Abu Dhabi has extended Microsoft 365 Copilot to roughly 35,000 government employees as part of its sovereign AI strategy.
Regulation is moving in parallel rather than behind. The DIFC's Regulation 10 already sets obligations for autonomous systems processing personal data in Dubai's financial free zone, covered in DIFC Regulation 10. The pattern across the Gulf is consistent: the deadline arrives before the tooling is mature, and organisations buy their way to compliance.
What's Next
The first real test is measurable. Entities that entered the 90-day cycle at the June workshop are due to have launched a redesigned service around mid-September 2026, which will be the first public evidence of whether the 2028 target is a plan or an aspiration. Watch for which services go first — high-volume, rule-bound approvals such as permits, licences and registrations are the natural starting point, and they are also the ones where an agent's failure is most visible to citizens.
For private-sector organisations in the Gulf, government procurement sets the standard the market then adopts. A federal definition of what "agentic" means in practice — task classification, human oversight thresholds, cross-entity data access — will become the reference architecture enterprises are measured against, whether or not they sell to government.
Two practical reads if you are planning against this: our AI agent readiness playbook for MENA SMEs on scoping a first deployment, and the enterprise agent sprawl governance crisis on what happens when the pilots multiply faster than the controls.
If you are mapping a first agent deployment onto systems you already run, talk to us — the honest first step is an integration audit of what your agents would need to read and write, not a model selection.
Source: UAE Cabinet and WAM (Emirates News Agency)