Abu Dhabi's Inception42 plans to begin deploying agentic AI systems across the UAE federal government later this year, with a national go-live targeted for the third quarter. The initiative, first announced in May, aims to have AI agents handling half of all federal government operations within two years, which would be the world's first attempt at agentic AI adoption at a national scale.
The rollout sits inside a broader collaboration between Inception42's parent company, G42, and Microsoft, expected to bring $15.2 billion in AI investment to the UAE. Under the arrangement, the companies' platforms are being integrated so that sovereign AI agents can scale across organizations while keeping data processing in-country.
Governments and businesses will build and manage agents through Inception42's Catalyst platform, while employees interact with them through familiar tools like Microsoft 365 Copilot, and agents built on either platform can run interchangeably on the other, removing the need to build the same tools twice across separate systems.
Unlike chatbots that respond to individual queries, agentic AI is built to carry out multi-step tasks independently, making decisions and completing work with limited human intervention. Koshy described the effort as a live, at-scale deployment running against a fixed timeline, framing the coming months as a genuine test of whether agentic systems can operate reliably inside government workflows rather than just in pilot environments.
Amr Kamel, General Manager, Microsoft UAE, stated that organisations are looking for AI tools that are not only powerful but also secure and easy to use.
“By enabling AI assistants to work with data that remains in-country and integrating them into the Microsoft tools people already use every day, we are helping governments and businesses innovate while maintaining security, compliance and control,” he said.
Koshy said that the harder problem organizations face isn't creating AI agents but integrating them, noting that many companies rushed into pilots that ended up siloed and disconnected, making it difficult to demonstrate any real return on investment. To address this, Catalyst includes a built-in "command center" that lets humans oversee agents, whether built in Copilot or Catalyst, and monitor failure rates, ensuring a person always reviews and directs the system so accountability remains clearly with individuals rather than the AI itself.
Also Read: Anthropic Rolls Out Claude Tag: AI Agent That Joins Slack Conversations, Tracks Context and Works Across Teams 24/7
Koshy pushed back on the idea that agentic AI deployment should be mistake-free from the start, arguing that expecting zero errors during this stage of experimentation isn't realistic. He said the real measure of success is how quickly the system learns from mistakes and scales transparently as it goes, a framing that positions the UAE's rollout as an iterative, closely monitored rollout rather than a finished, risk-free product.
The UAE's planned rollout of agentic AI marks an ambitious step toward integrating autonomous AI into public services at a national scale. If successful, the initiative could redefine how governments use AI for operations while highlighting the importance of secure integration, human oversight, and continuous learning in large-scale AI deployment.