Abu Dhabi aims to be an AI-native government by 2027, Lootah says

Abu Dhabi is aiming to become what it calls the world’s first AI-native government by 2027. H.E. Wesam Lootah, Director General of GovDigital at the Department of Government Enablement, has set out why the outcome depends on data, resilience and keeping people accountable for the decisions that matter. The statements appear in interviews and briefings through September and October 2026. No independent audit of the claimed use-case count or transaction volumes was published alongside them. GovDigital sits inside the Department of Government Enablement. It runs the shared digital architecture for Abu Dhabi government. Lootah has said more than 100 AI use cases are already live across more than 40 government entities, with a further pipeline of more than 200. The unified platform TAMM brings together more than 1,170 services. It is used by more than 4.5 million people and processes more than 17 million transactions a year, 96 percent of them completed without a human in the loop. By the end of 2025, 95 percent of public-sector employees had completed AI training. The work sits inside the Abu Dhabi Government Digital Strategy 2025–2027, which carries an AED 13 billion allocation. For a resident in Abu Dhabi the useful test is what happens when something goes wrong. A system that renews a trade licence or books a routine appointment without being asked can save time. The same system that declines a benefit or flags a household still needs a person who can reverse it. Lootah’s emphasis on keeping people accountable for the decisions that matter is the recognition of that line. The statements do not list which decisions have been classified as requiring a human, or how a resident appeals one. Data is the other dependency he names. Shared platforms and sovereign infrastructure are the stated foundation. A model that sees across entities can anticipate a need. It can also propagate an error across entities. Resilience, in this framing, is the ability to keep services running when a model or a data feed fails. The public comments do not give a measured uptime figure or a published incident count. The training number is large. 95 percent of employees completing a course is a baseline of exposure. It is not a measure of how often they override a model or how often they accept one. A workforce that has sat through the modules can still defer to the output. Accountability is a practice, not a completion rate. Abu Dhabi has been building the shared layer for years. TAMM, AutoGov for routine renewals, and the sovereign cloud are already in use. The 2027 target is the point at which the government says the redesign will be complete. Readers can check that claim against the published counts: use cases live, transactions autonomous, employees trained. They cannot yet check the unpublished ones: error rates, appeal volumes, or the list of decisions reserved for a person. A further limit is what “AI-native” means in daily use. A resident who renews a licence through TAMM and never sees a form has experienced one version. A resident who is told a model has flagged an application and must visit a centre has experienced another. The public statements do not separate the two. They report the aggregate: use cases, transactions, training. They do not report the share of cases that still require a person at the end. The AED 13 billion figure is the envelope for the digital strategy. It covers infrastructure, cybersecurity, governance and workforce programmes. It is not a line item for any single model. A resident cannot see how much of it sits in the systems that touch their file. The published number is the total. The split is not. A clinic that receives a real-time clinical context for a doctor and a road system that flags a hazard are both counted in the 100. They are different kinds of decision. One may be reversible in the room. The other may already have been acted on. The aggregate does not distinguish them. If the Department of Government Enablement publishes the decision categories that stay with people and a measured error or appeal rate, the accountability claim has a record. If the next line is only another use-case total, count the cases a resident can reverse. The figure missing is both.