Agentic AI has moved firmly into large-scale production across the UAE, according to new global research from Dataiku. Conducted by Harris Poll, the research shows that 62% of UAE CIOs estimate that more than 50 AI agents are actively running in production within their organisation, while 15% report operating more than 500, the highest proportion across the eight markets surveyed, and substantially higher than the 9% global average.
UAE organisations are also investing in the expertise required to support this expansion. Among CIOs whose organisations are creating or planning embedded roles for AI experts or forward engineers, 22% expect to recruit more than ten people over the coming year, the highest compared to other countries in the study.
Encouragingly, the scale of deployment has not been associated with greater reported harm. While 80% of UAE CIOs have encountered an agent that violated business intent, policy, or expectations, while continuing to operate within its technical parameters, most say the behaviour was caught before it caused damage. Almost two-thirds (62%) report that the incident was caught before causing customer, financial, compliance, or operational impact, the highest proportion globally, and compared with 48% overall. Only 18% say such an incident resulted in actual impact, the lowest figure across the markets surveyed and substantially below the 31% global result.
UAE organisations are also planning for disruption in the technologies underpinning their agent strategies. Almost two-thirds (63%) plan to use more, or different, models to strengthen business and AI continuity, while 62% are considering open-source alternatives as a hedge against future unavailability. These measures could help organisations build AI sovereignty and insulate themselves against real-world disruptions such as the temporary suspension of Anthropic’s Fable 5 and Mythos 5 models under a U.S. government order.
However, the research shows that preventing immediate harm and containing a problematic agent across the enterprise are very different capabilities. Only 5% of UAE CIOs say their organisation could reliably identify and contain a problematic agent across environments within one to two hours, among the lowest results in the study and half the worldwide figure of 10%. A further 27% could intervene later that day, meaning more than two-thirds could not reliably contain the agent on the same day. For 45%, containment would take between 24 and 48 hours, and almost one-quarter (23%) say they would require two days or more, one of the highest figures in the study.
“CIOs in the UAE are running agents at a scale few markets have reached, and they’re unusually good at catching bad behaviour before it does damage. But detection and control are different capabilities, and the research shows a real gap between them. An agent you can spot in one system but can’t stop everywhere else isn’t contained — it’s just been noticed,” said Florian Douetteau, CEO and co-founder of Dataiku.
The containment gap comes as AI agents increasingly operate across different platforms, data environments, and business applications. Only 18% of UAE CIOs report having a fully unified view of agents across these environments, while just 12% say agent lifecycle management is standardised throughout their organisation. Without that unified view and standardised lifecycle, an agent detected in one environment could continue operating elsewhere before teams are able to intervene.
At the same time, pressure on UAE technology leaders to deliver measurable results is intensifying, with 100% of UAE CIOs surveyed reporting greater board pressure to demonstrate measurable AI ROI compared with 2025. Meanwhile, 93% say their CEO has indicated that their job security depends on AI outcomes. Against this backdrop, 90% would stake their job on delivering measurable AI results.
“The UAE has established a strong foundation for agentic AI, with enterprises deploying at scale, investing in specialist expertise, and preventing a high proportion of agent incidents from causing real-world harm,” said Sid Bhatia, Area Vice President and General Manager, Middle East, Turkey, and Africa at Dataiku. “The next stage is to connect those safeguards across the enterprise. That means giving CIOs a unified view of their agent population, clarity on who has the authority to act, and the ability to intervene immediately, regardless of where an agent was built or which system it uses. Rapid containment will be critical to protecting the UAE’s progress as agentic AI becomes embedded across the economy. For CIOs, there is also a significant personal stake in getting this right.”
