Today's Edition
SARA ALZAABI (ABU DHABI)
As the UAE reviewed the next phase of its Agentic AI Project, experts pointed to cybersecurity, trusted data, human oversight, workforce skills and responsible adoption as key priorties.
The Agentic AI Project is a unified national framework aimed at expanding the use of Agentic AI across government to improve efficiency, service quality and speed of execution.
At the centre of the project is FedAI, a national technical ecosystem designed to enable federal entities to develop, adopt and operate Agentic AI solutions within a flexible and secure environment.
The project also includes a customer experience lab and initiatives to develop government AI solutions.
Its first phase focused on building capabilities, assessing services and operations, and identifying priorities, with selected projects expected to move into implementation.
Aletihad spoke to experts in technology, cybersecurity, AI and business about the wider requirements for increasingly autonomous AI systems.
Cybersecurity Challenge
On securing AI from government cloud to citizen device, Debo Zhang, CEO of HONOR GCC, said moving from reactive chatbots to AI systems capable of taking action significantly changes the cybersecurity challenge.
"As AI transitions from reactive chatbots to autonomous agentic execution, the cybersecurity attack surface expands significantly. Malicious actors will increasingly utilise adversarial attacks and deepfakes to manipulate these agents," he said.
Zhang argued that securing FedAI and government infrastructure alone would not be enough, as the devices citizens use to access AI-enabled services also form part of the security chain.
"A highly secure government AI network is only as resilient as the personal devices citizens use to interact with it," he said.
This becomes particularly important when AI agents are authorised to handle sensitive information or execute transactions.
"While routine administrative tasks can and should be fully automated, any high-stakes transaction involving financial payments, legal rights, or sensitive personal data must maintain a strict 'human-in-the-loop' safeguard." Zhang said the combination of shared infrastructure and common governance could ultimately become an advantage for the UAE.
"The combination of FedAI's unified infrastructure and standardised Agentic AI governance positions the UAE to secure a massive global competitive advantage in digital governance." This focus on securing increasingly autonomous systems extends beyond government.
AI for Cyber Defence
As businesses also become more dependent on AI, Shoaib Yousuf, Managing Director & Partner at BCG, said cybersecurity would need to move beyond the remit of technical teams. "UAE organisations must prioritise elevating cybersecurity from a technical concern to a board-level strategic priority," he said.
BCG research cited by Yousuf found that more than 70% of organisations in the Middle East had faced suspected AI-enabled cyberattacks in the past year.
Meanwhile, 70% of companies in the region were prioritising AI for cyber defence.
AI, he said, was therefore creating both a threat and an opportunity. "We are witnessing a duality in how AI is reshaping cybersecurity."
On the defensive side, AI-powered systems are increasingly being used for threat detection, incident response, anomaly identification and prioritising security alerts.
"The region's vision and willingness to prioritise advanced solutions at a rate that outpaces the rest of the world demonstrates a mature understanding that defence must evolve at the same speed as the threats organisations face."
Yousuf said human expertise would, nevertheless, remain essential.
BCG's research found 64% of Middle East organisations were seeking specialised cybersecurity professionals.
"The UAE organisations that seize this opportunity, leveraging AI defensively while remaining vigilant to its risks, will set the standard for cyber resilience in the region and beyond."
As AI takes on more tasks and executes workflows, the focus is increasingly shifting to which decisions should remain in human hands.
Human Responsibilities
AI and data expert Vasudha Khandeparkar said the distinction would become increasingly important as AI agents move from providing assistance to executing entire workflows.
"The question is shifting from which jobs stay human to which decisions should," she said. "AI agents are becoming increasingly capable of executing workflows, but accountability, ethical judgment and navigating ambiguity remain fundamentally human responsibilities."
Rather than simply automating existing tasks, Khandeparkar said organisations should rethink how work is divided between humans and machines.
"AI should manage operational execution, while people focus on setting direction, managing exceptions, building relationships and making decisions where context matters more than efficiency."
She said organisations best prepared for the next stage of AI would not necessarily be those deploying the most sophisticated technology. "Future-ready organisations won't necessarily be those with the most advanced AI, but those that integrate it most effectively into how work gets done."
That, she said, would require strong data foundations, governance, transparency, oversight and employees capable of working effectively alongside increasingly autonomous systems.
The impact of the UAE’s wider AI transformation is also extending beyond government. As advanced AI capabilities become more accessible, smaller businesses are gaining access to tools once largely available to bigger companies.
Reliable Data Foundations
Hetarth Patel, an AI for SMEs expert (Vice President MEA, Americas & Asia Pacific at WebEngage), said technologies that once required the resources of major companies were increasingly within reach of much smaller organisations.
"Capabilities that once needed enterprise budgets, behavioural segmentation, journeys that run themselves, are now within reach of a fifty-person company," he said.
But access to AI alone will not determine which businesses succeed.
"Every SME will have access to the same models within a year or two, so adoption alone stops being interesting. It will come down to discipline." Patel said businesses should establish reliable data foundations before deploying AI and ensure that customer trust develops alongside their technology.
"Start with data before touching any AI tool. Most SMEs have customer information scattered across half a dozen systems, and no model performs well on fragmented inputs."
As AI becomes embedded across both public and private-sector services, he said transparency and responsible data use could themselves become competitive advantages.
"The winners will be the businesses that treat consent and transparency as primary design principles from day one. It is crucial to focus on building trust alongside your tech stack."
