Huawei’s KOH to Sabq: Data, Trusted Models and Talent Underpin the Shift to Agentic Cities
Huawei’s Global Chief Public Services Industry Scientist Hong-Eng KOH told Sabq at LEAP that building agentic cities relies on three foundations: shared, well-governed data; trusted, locally adapted AI models; and people trained to develop and use these systems. He highlighted Riyadh’s potential, st…
The transition to agentic cities depends on three foundations: accessible and shared data, trusted AI models adapted to local needs, and people equipped to develop and use the technology, according to Hong-Eng KOH, Global Chief Public Services Industry Scientist of Huawei Technologies Co., Ltd.
Speaking to Sabq on the sidelines of LEAP, KOH described an agentic city as an environment where multiple AI agents work together to complete tasks across agencies. Rather than simply generating an answer or triggering a single action, these agents can use different applications, tools and models to carry out a broader mission.
He illustrated the concept with a hazardous materials accident, where agents responsible for traffic, hospitals and public alerts could coordinate their responses. The same approach could extend to municipal services, he said, including restaurant licensing, which involves several authorities and could become more seamless through coordinated agents operating with limited human supervision.
Asked how many agents a city such as Riyadh would need, KOH said the answer depends on the intended applications and the availability of data. He pointed to opportunities across education, healthcare, city administration and crisis management. In education, agents could support teachers with curriculum preparation, assessment questions and evaluation, alongside efforts to teach students how to use AI.
Technology alone would not ensure success, he cautioned. Bureaucracy and reluctance to share data across agencies could obstruct progress, making data governance as important as the underlying systems. He also stressed the need to protect data quality, security, privacy and sovereignty, particularly when information is submitted to AI services hosted by external providers.
On AI models, KOH advocated adapting foundation models to Saudi values, languages and local requirements, referring to ALLaM in his discussion of sovereign models. Developing AI engineers and enabling users to apply the technology effectively were equally important, he added.
The technical infrastructure, he explained, spans platforms for developing and running agents, systems for model training and inference, and interconnected computing, storage and networking resources. He highlighted persistent memory as a requirement for agents to retain context over time and provide more relevant assistance to their users.
KOH said earlier adoption of these technologies, combined with investment in engineering skills and computing capacity, could support the goals of Saudi Vision 2030.