Thai Officials Warn: Fix Real Problems First, Not AI Hype

 

 

Government and industry leaders in Bangkok say Thailand's AI push must be built on data sovereignty and practical use cases — not headline-grabbing technology purchases.

 

Government and industry leaders gathered here to confront a hurdle facing Thailand’s AI ambitions: how to scale artificial intelligence without ceding control of the country’s data or wasting public money on the wrong problems.

 

The panel, “Successfully Enabling Intelligence Thailand", was held during the Huawei Thailand Digital & AI Summit 2026 at the Queen Sirikit National Convention Center, moderated by Dr Chai Wutiwiwatchai, executive director of the National Electronics and Computer Technology Center (NECTEC).

 

It brought together Charnkrit Triratananont, director general of the National Village and Urban Community Fund Office; Paisan Khonjumpa, deputy secretary-general of the National Institute for Emergency Medicine; a representative from the Bureau of the Budget; and Hong-Eng Koh, Huawei’s global chief public service industry scientist.

 

 

Start With the Problem, Not the AI

Asked where agencies should begin, Charnkrit was blunt: “Look at problem first, not AI first.” He said organisations need to analyse pain points at both the user level and, ultimately, the level of the citizens they serve, then chase small, quick wins rather than sweeping projects.

 

“When we start thinking about 'big wins', it often turns into a 'big fail'," he said.

 

Koh backed that approach with lessons from two decades of government-technology projects worldwide. Agencies routinely rush in without an overall plan, he said, buying hardware before knowing what problem it solves.

 

 

 

"You need to plan big... but you need to start small,” he said, adding that government is unlike a bank or corporation, where leadership can stay in place for decades: with elections roughly every four years, agencies must show citizens quick, visible value or face backlash over spending.

 

Koh also flagged a factor he said gets overlooked amid the AI conversation: connectivity. Huawei’s RuralStar technology, a low-power repeater built for areas without existing infrastructure, exists precisely because data collection and AI training are impossible without a network to move that data across.

 

 

The Real Obstacles

Pressed on what stands in the way, panellists pointed less to technology than to people and process.

 

Paisan cited two barriers specific to emergency medicine: data quality, since AI systems still depend on clean, linked information even when a physician’s diagnosis ultimately isn’t; and professional trust, noting some clinicians reject AI tools outright before ever using them, a hesitation he said takes time to overcome.

 

He said the deeper issue is coordination — individual staff already experiment with AI for basic tasks like drafting slides, but that doesn’t add up to an organisational strategy.

 

“Each one is not yet AI for the organisation," Paisan stated, arguing agencies need someone to gather scattered use cases into a shared architecture before AI delivers real value.
 

 

 

 

Charnkrit went furthest, laying out five recurring obstacles: AI use cases that don’t serve an institution’s core mission; siloed deployments with no shared framework to scale from; scattered data that AI can’t connect; workflows that never actually change, which he said reduces the value of any AI system to zero; and a lack of clear ownership at the organisational level.

 

“The toughest asset to manage is human,” he said. “That’s the problem.”

 

 

Sovereignty Has Three Layers

Koh’s most detailed remarks concerned data itself. Technologies come and go, he said, but data is the one foundation that keeps compounding in value – echoing a framework he attributed to Nvidia chief executive Jensen Huang, who has described AI’s supply chain as four layers: energy, GPU infrastructure, models, and data-dependent applications.

 

“With no data, there is no model,” Koh said. “You cannot fine-tune.”

 

He broke sovereignty into three distinct concepts.

 

Digital sovereignty means a country must be able to control its technology if a supplier cuts it off — citing Thailand’s stated goal of drawing 30% of GDP from the digital economy by 2030 and noting that China’s early push to build domestic technology now insulates it from US leverage.

 

Data sovereignty, a term Koh said the European Union coined roughly eight years ago, addresses the U.S. CLOUD Act (Clarifying Lawful Overseas Use of Data), which can compel American companies to hand over data to U.S. authorities even when it’s stored abroad — meaning using a U.S. cloud provider in Bangkok doesn’t fully shield data from Washington’s reach.

 

AI sovereignty, the third layer, concerns keeping language models free of foreign values or bias, which Koh said is why he was glad to see homegrown efforts like ThaiLLM taking shape.

 

 

Balancing Trust and Innovation

Asked how to balance innovation against these risks, Paisan said trust matters most where it affects citizens directly, such as e-commerce pricing – but noted his own agency’s medical data rarely leave Thailand, so he feels less exposure to the issue personally.

 

NECTEC’s Dr Chai described managing the balance through access controls—assigning permissions by department and keeping sensitive data out of AI systems entirely to lower the risk of leaks.

 

Charnkrit argued sovereignty and innovation “should walk together", built on trust from day one. In practice, he said, that means grading data by sensitivity and running a hybrid model — some workloads in the cloud and others on-premise — while accepting that Thailand can’t build everything itself.

 

With the US and China leading AI development, he said, the realistic path is developing domestic capability while adopting from established leaders, rather than pursuing total self-sufficiency.

 

 

What Success Requires

Closing the panel, each speaker named a different priority. For Chai, it’s leadership: once executives commit to a use case, the budget follows. For Paisan, it’s cost discipline — optimising token consumption so AI systems don’t become an unsustainable drain on public budgets.

 

For Charnkrit, it’s people: “Artificial intelligence is here, but humans are real intelligence,” he said, arguing AI’s purpose is to make people smarter, not replace the work of developing them.
 

Asked about Huawei’s own commitment, Koh pointed to the company’s 28 years operating in Thailand; more than 210,000 employees worldwide — over half in R&D, against an industry-typical 20%; and eight consecutive years as the world’s top filer of international patents, according to the World Intellectual Property Organization.

 

He said Huawei builds full-stack, open-standard systems rather than single products and relies heavily on local staff: 80% of its international workforce is non-Chinese, and in Thailand the company works alongside more than 2,000 local partners.

 

The panel’s throughline, across four very different agencies, was consistent: Thailand’s AI transformation will be judged not by how much technology gets bought but by how well it’s grounded in real problems, trustworthy data and people equipped to use it.