AIBRIEF
The UAE wants half of government work handled by autonomous AI agents — software that takes multi step actions across other systems — within two years. The harder deadline is whether the data and audit stack underneath can keep up.
Agentic AI, systems that take multi-step actions across other software rather than chatbots that only answer, moved from slide to schedule at the UAE's federal Identity Authority on 11 August. The Authority convened an Agentic AI Forum in Abu Dhabi to start building toward a Cabinet target: half of government operations running on agentic AI within two years, alongside a national program to train 80,000 federal employees.
The Cabinet set the 50% target on 23 April 2026 and approved the training program on 18 May 2026. Major General Suhail Saeed Al Khaili said the forum responds to UAE leadership directives and a strategic national project launched in April.
The two-year clock puts weight on the prerequisites Deloitte Middle East partners Mike Horseman and Shargil Ahmad said in commentary carried by Gulf News are often skipped: connected data, secure integration, identity management, data sovereignty, continuous monitoring, and explainable, auditable agent actions. Horseman argued high-risk decisions should remain under meaningful human oversight, with agents granted only the autonomy a specific task requires.
Ahmad said officials will continue setting policy and stay accountable, while AI orchestration, human-centred design and change leadership grow more important. The open question is whether the data and audit stack can be built fast enough to let an agent carry a citizen's request across multiple systems without a caseworker stitching it together.