Exclusive: Inside Singapore’s AI push with Temus’ Matt Johnson

Deploying AI in the real world requires more than hype and hope - it demands strategy, caution and, above all, learning from experience.

That's the message from Matt Johnson, Managing Director for AI and Data at Temus, who believes the key to meaningful AI adoption lies in matching ambition with maturity.

"A lot of people are testing solutions," he explained during a recent interview. "What differentiates the successful ones is a clear understanding of the technology's strengths and weaknesses - and picking the right applications."

Johnson is sceptical of the belief that simply prompting a large language model is enough to unlock business transformation.

"There's a lot of belief that prompting is all you need," he said. "Take a large language model, give it a prompt, and expect it to deliver expert results. But there's very little difference between a prompt and the welcome pack you're given on your first day at a job."

He compared effective AI deployment to onboarding a new employee: the real skill doesn't come from the manual - it comes from experience.

"In a human context, you improve by interacting with the company. In AI, that experience is data. That's how you actually build expertise."

For Johnson, that means moving away from isolated pilots or generic deployments and towards long-term learning cycles. "The most successful applications are based on learning from real-world use. That means monitoring, tuning, and continuously improving."

He's blunt about the limits of so-called "AI" when human input is still doing all the learning.

"Prompt engineering is not machine learning," he said. "The machine isn't learning. The human is tweaking the prompt, running the model, and changing it again. That's not sustainable."

Instead, Johnson sees the future in domain-specific, fine-tuned systems that specialise in particular tasks - just as people do.

"We're moving away from general models and prompts, and towards models that are trained to be experts in specific domains," he explained. "That's where the real value is."

A strong example lies in Temus' work in insurance, where Johnson's team is helping sales agents train through AI-powered role play.

"It's a high-impact use case with manageable risk," he said. "We start by supporting the training of humans—helping agents practise face-to-face sales - then build towards AI tools that support those same interactions."

The strategic sequence matters. By starting in lower-risk areas like training, companies can collect the right datasets, build trust in the technology, and eventually move towards AI tools that assist, or even lead, customer interactions.

"It's a roadmap," Johnson explained. "You start with use cases that work today, prove the value, and then move forward—cautiously but purposefully. That's how you get from where AI is to where it needs to be."

One field that demonstrates the potential of AI/data integration is medicine. Temus is supporting Singapore's national precision medicine efforts through its work with Precise, the Ministry of Health body overseeing the SG100K genome project. While the scientific discovery lies elsewhere, Temus is building and managing the data pipelines that enable it.

"These are enormously large datasets," he said. "It's genomic data - gigabytes per person, multiplied by 100,000. Our engineers are helping make that data usable."

For a small country, Singapore punches above its weight, Johnson said. OpenAI's own usage statistics show Singapore has the highest per capita use of ChatGPT in the world, and the company recently chose the country as its regional office hub. Temus was there at the opening.

"They were surprised at the level of AI adoption we've seen here," Johnson said. "Some of the deployments we described were ahead of what's happening in Silicon Valley."

Singapore's advantage, in his view, is the strength of its public-private partnerships. "

There's real collaboration between research, government and business," he said. "Whether it's EDB, the Digital Industry Singapore group, or AI Singapore, there's a shared strategy. And that's rare."

He's especially proud of Temus' relationship with AI Singapore, which has trained a significant portion of the firm's AI engineers. "About a third of our engineers come from that programme," he said. "They've been essential in developing talent for the nation."

With an ageing population and a small workforce, Singapore has pragmatic reasons to invest heavily in AI, Johnson noted. "The question for Singapore is: how do we stay productive as our population ages? AI is part of the answer."

But while the stakes are high, Johnson is also clear-eyed about what AI can - and can't - do. "There's a narrative out there that AI will solve world hunger," he said. "The truth is somewhere in between."

He pointed to medicine, again, as a more realistic frontier. "If you look at what Demis Hassabis and Google DeepMind are doing - using AI for drug discovery, or to trigger immune responses against cancer - those are real breakthroughs being worked on now. They're serious, and potentially life-changing."

Temus might not have Google's scale, but Johnson believes Singaporean firms can still play a role. "We're not doing the science," he said. "But we are helping to prepare the data that will enable it. And that's just as critical."

Looking ahead, Johnson remains optimistic - but cautious - about what's next. "There's a lot of aspiration out there," he said. "Our job is to make sure that aspiration translates into productive, safe, and valuable deployments."

He added: "That's how AI delivers - not with flashy demos, but with systems that learn, adapt, and improve over time."