AI Leaders Aren't the Ones with Access — They Redesign Work

 

Enterprise leaders at a Singapore-China forum say the AI divide is organisational, not technological — a message with direct relevance for Thai firms.

 

For Thai companies still experimenting with AI chatbots without seeing much return, a panel of enterprise leaders in Singapore this week offered a pointed diagnosis: the technology was never the hard part. The hard part is whether leadership is willing to redesign how work actually gets done.

 

The discussion, titled "Putting AI to Work: What Separates the Enterprise Leaders", formed the second panel of the 17th FutureChina Global Forum, held on Thursday (September 24) in Singapore under the theme "Strengthening Resilience, Rebuilding Trust".

 

The two-day event, organised by Business China — a non-profit founded by Singapore's first prime minister, Lee Kuan Yew — brings together business and government figures from China and Singapore, and this year devoted its opening day entirely to artificial intelligence for the first time.

 

Moderated by Kuek Yu Chuang, deputy chief executive of SPH Media, the panel featured Cynthia Zhang, founder of venture firm FutureX Capital; Guan Dian, co-founder and Asia-Pacific general manager of intellectual-property intelligence firm Patsnap; Simon See, chief solution architect and global head of Nvidia's AI Technology Center; Thomas Wee, chief executive of Gleneagles Hospital Singapore and head of the AI coordinating team at IHH Healthcare Singapore; and Zhou Yuxiang, founder and chief executive of manufacturing software firm Black Lake Technologies.

 

 

 

 

The dividing line is organisational, not technical

According to the forum's briefing notes, Singapore's "AI gap" is increasingly an enterprise problem rather than a technology one: employees already experiment freely with AI tools, but companies remain stuck at the pilot stage because of integration costs, data-security worries and internal resistance to changing established workflows.

 

The real dividing line, organisers noted, is whether leadership is prepared to redesign work around AI rather than simply hand staff a chatbot. That theme was reinforced repeatedly on stage.

 

Zhou Yuxiang described how his firm shifted away from selling AI as a plug-in digital tool and instead built narrow "agents" for specific factory tasks, such as converting engineering design files directly into work instructions — traditionally a job requiring skilled human planners. The first version worked only 60 per cent of the time.

 

 

Improvement came not from a better model but from feeding it clean, proprietary factory data and, crucially, capturing corrections made by human supervisors once the system went live.

 

"Once you reach a usable level, your agent will be embedded in the users, and the magic happened when I observed my agent running at the client's site for one to two months—the accuracy level improved from 92 per cent steady to 97 and 98 per cent," Zhou said, describing the process informally as "reinforcement learning from the engineer" rather than from the machine alone.

 

Guan Dian, whose company supports research-intensive clients such as Huawei, Dyson and Tesla with patent and innovation intelligence, made a similar point about the gap between model capability and real-world adoption.

 

 

 

 

"Diffusion is not as quickly as the development of [model] capability," she said, arguing that companies like hers play a critical role in translating raw AI power into usable, industry-specific tools — a role she believes remains underfilled across most sectors.

 

Where healthcare draws the line For sectors handling sensitive data or safety-critical decisions, the panel underlined that human oversight cannot simply be automated away.

 

Thomas Wee described how IHH Healthcare Singapore has organised its AI use into three tiers: clinical tools that directly affect diagnosis, such as software that flags tumours during endoscopy procedures more reliably than the human eye; "clinical support" tools such as AI scribing, which lets doctors focus on patients instead of typing notes; and operational tools, including an AI-generated nurse rostering system.

 

 

 

That rostering tool, Wee said, lets nurses submit shift preferences — such as needing to collect children from childcare — which the system then aggregates into a roster acceptable to the whole team.

 

"It's causing a lot of anxiety among our nurses, and at some point a lot of attrition," he said of the old manual system, adding that after roughly 18 months of use, "our attrition rate has gone down tremendously."

 

Yet Wee was equally firm about limits: Singapore's medical regulations require a qualified practitioner to remain present even when AI tools assist in procedures.

 

"I cannot use the AI to just look at the scope, and the surgeon goes and drinks coffee," he said, describing patient safety and data confidentiality as the sector's "topmost priority" even as the hospital group pushes to adopt promising tools from anywhere in the world, including China.

 

 

 

The human bottleneck, not the model

Cynthia Zhang, whose firm has backed more than 1,700 AI-native companies since 2018, offered a candid admission from the investment side: despite trying to be "the most AI-native VC in the world", her team has found that judging founders and building relationships—the core of venture decision-making—still cannot be outsourced to AI.

 

"AI makes us much busier, probably five times more busy than like two years ago," she said, while confirming the return on that investment has been positive.

 

Simon See of Nvidia framed the broader challenge as a pace mismatch: AI capability is advancing "not linear, not polynomial — it's exponential," he said, but adoption depends on human decision-makers who cannot move nearly as fast, generating what he called "a lake of anxiety" inside organisations trying to keep up.

 

 

A lesson for Thailand and ASEAN

The panel's clearest message for markets like Thailand may have been about regulatory posture. See contrasted China’s experimental culture—where companies are “willing to experiment, try, collect the data and then try again if it fails”—with Singapore’s more cautious instincts.

 

He likened the latter to pedestrians waiting for a green light at an empty crossing. However, he argued this caution makes Singapore, and the wider region, ideal partners for collaborating with China on applied AI and data use, rather than competing to build the biggest models.

 

He also pointed to highly practical and transferable applications. For instance, See observed AI-monitored vertical farming in China that tracks plant health, UV exposure, and nutrient dosing.

 

He suggested this technology directly supports Singapore's food security goals and could be seamlessly adapted for Thailand's agricultural sector.

 

Zhang highlighted another structural shift impacting the region. She cited projections that China will eventually have around 100 million engineers in AI-related fields.

 

Combined with cheaper electricity and robust supply chains, these advantages are already driving Chinese AI firms to set up regional headquarters in Singapore.

 

Global adoption of Chinese AI models has surged from roughly 2 per cent to over 50 per cent in just a year. As a trusted, neutral hub, Singapore is perfectly positioned to help these companies serve global markets.

 

For Thai policymakers watching this dynamic, the lesson is clear. Rather than racing to build frontier models, the focus should be on practical infrastructure.

 

This means securing cheaper power, building engineering talent pipelines, and creating regulatory sandboxes. These safe zones would allow targeted, well-supervised AI experiments in healthcare, agriculture, and manufacturing, avoiding the risks of unchecked deployment.

 

 

Asked for one closing message each, the panellists converged on urgency tempered by caution.

 

"I don't think the Google moment for AI has happened yet," said Guan Dian.

 

See urged organisations to adopt an "institutional attitude" of experimentation: "You need to fail early... if you do not try, then you will not gain the experience."

 

Wee argued healthcare's most guarded areas were precisely where "the real breakthrough" would come. Zhang was blunter still, urging every business leader in the room to "register, download open-source models, put them into your company computer" and start using AI as a genuine decision-making tool rather than a side experiment.