Inception42 targets UAE agentic AI rollout in third quarter

• To automate half of government tasks
• Limited human role in decisions
• Part of collaboration with Microsoft

Abu Dhabi company Inception42 plans later this year to begin deploying AI systems capable of carrying out complex tasks across the UAE administration, with the aim of automating half of federal government operations within two years.

Known as agentic artificial intelligence, the technology goes beyond chatbots by carrying out multi-step tasks independently, making decisions and completing work with limited human intervention.

Inception42, the sovereign agentic AI company owned by Abu Dhabi technology holding group G42, expects the national rollout to go live by the third quarter of this year, the company’s chief executive told AGBI.

The initiative, first announced in May, aims to have AI agents handling 50 percent of federal government operations within two years – positioning the UAE for the world’s first attempt at agentic AI adoption on a national scale.

“There is a clock that has started ticking, and people have it on their dashboards,” said Ashish Koshy, CEO of Inception42. “We’re doing real-life deployment, and at scale, on a clock.”

As part of a wider collaboration between G42 and Microsoft, which is expected to bring $15.2 billion in AI investment to the UAE, the two companies unveiled a partnership designed to let organisations move away from disconnected AI experiments and towards a single, governed system of agents that works across an entire enterprise or government body.

Under the arrangement, the two companies’ platforms are being woven together so that sovereign AI agents can scale across organisations while keeping data processing in-country.

Governments and businesses will build and manage their AI agents using Inception42’s software platform, Catalyst.

Employees will access the AI agents through familiar workplace software such as Microsoft 365 Copilot, allowing them to complete routine tasks and automate parts of their daily work.

AI agents created on Inception42’s Catalyst platform can be used directly within Copilot. Likewise, agents created in Copilot can be used in Catalyst.

This means governments and businesses need not build the same AI tools twice or run them on separate systems, making it easier to deploy and manage across different applications.

Integration challenge

Koshy said the biggest challenge organisations face isn’t building agents, it’s integrating them. Many companies rushed into AI pilots that ended up siloed and disconnected, making it hard for decision-makers to see any return on investment.

“We’re moving from a situation with multiple AI pilots that don’t talk to each other, to a unified deployment, making sure that whatever agents you build, you do it on one platform,” Koshy said, adding that a human always remains in the loop overseeing the system.

However, when AI makes decisions that could affect citizens and residents, who is accountable if an agent gets something wrong?

Koshy said Catalyst addresses this through a built-in “command centre” in which humans can oversee the agents assigned to them – whether those agents were built in Copilot or in Catalyst – and monitor how often the agents fail. Because a human is always reviewing and directing the agents, he said, accountability stays clearly with the individual, not the AI.

He also pushed back on the idea that agentic AI deployment should be mistake-free from the outset.

“If the expectation is that while we are experimenting with AI, we will not have mistakes made, I think that does not exist,” he said. “It’s how quick we can learn from those mistakes and scale that in a manner that we become transparent.”

Further reading:

• AI-driven ‘virtual lawyers’ to launch in UAE
• UAE’s AI rules ‘prioritise innovation but run risk of gaps’
• Why Arabisation is critical for local AI