AI Can't Make Whisky: Business Chiefs on Keeping the Human Touch at Work

 

Leaders of DBS, Suntory and Salesforce South Asia say upskilling, clear data ownership and human judgement must anchor AI adoption at work.

 

As companies pour billions of dollars into artificial intelligence, three Asian business leaders say the harder task is not the technology but the people and trust around it: teaching staff to build with AI, redeploying roles it replaces, and keeping human judgement in the loop.

 

Speaking on a CNN Global Perspectives panel on "Trust, Culture and the Future of Work", moderated by CNN's Will Ripley, the leaders of DBS Group, Suntory Holdings and Salesforce South Asia described how they are adopting AI while protecting the culture that holds their organisations together.

 

 

Learning by building

Tan Su Shan, chief executive of Singapore's DBS Group, said the bank has built about 40,000 AI agents, of which around 10,000 are actively used. The point, she said, was less the count than the learning.

 

At a leadership conference earlier this year, she handed everyone a T-shirt that read, "Talk is cheap. Show me your agent." The aim was to make it safe for staff to learn by doing.

 

"We're all off the same starting point at the same time," she said, "doesn't matter whether you're 80 or 18." The challenge ran from the chairman to the youngest analyst: build your own agent.

 

Staff do so inside what she called a walled garden, an in-house system running on the bank's own servers, with no data sent outside and all work grounded in governance and control frameworks.

 

"Let a thousand flowers bloom," she said, whether the agent helps analyse deposit rates or write credit memos.

 

 

 

From "pilot hell" to production

Arundhati Bhattacharya, president and chief executive of Salesforce South Asia, said many businesses were until recently stuck in "pilot hell": running trials they did not know how to scale.

 

Her advice is to pick one or two use cases where "the shoe pinches most", build them end to end in a live production environment, and measure the effect on the bottom line, top line and qualitative indicators such as market share or net promoter score.

 

If it works, iterate and move to the next. A small pilot, she argued, cannot prove value.

 

 

Redeploying, not just replacing

Asked how companies should support staff whose roles AI is phasing out, Bhattacharya cited Salesforce's own help line for customers who have already implemented its products.

 

It was once staffed entirely by engineers. After the work was handed to AI agents, she said, the line achieves 84 per cent first-call resolution, and only about 2 per cent of queries are still passed to human agents. \

 

About 2,000 engineers were upskilled and redeployed as "forward deployed engineers", who go to customers' sites to help them build agents. Tan offered a similar example.

 

The most senior union representative at DBS became an "agent conductor", evaluating and directing AI agents, she said. Bank tellers, rarely needed as cash use declines, now work on video ATMs or sit with customers as bankers.

 

 

 

Hire for attitude, pair old with new

For people starting out, and those who hire them, Tan said attitude counts for more than experience. Her managers, she said, would rather have an intern with no experience but a great attitude, armed with AI, than an experienced middle manager with a bad one.

 

"Attitude matters in this time of AI. Agility matters," she said, adding that the issue is not age but the ability to unlearn and relearn.

 

With CVs increasingly written by AI and screened by AI, she advises managers to bring candidates in as interns first and watch how they handle a crisis.

 

DBS also pairs veteran staff who hold deep domain knowledge with young, AI-literate analysts and asks them to build an agent together. The approach produced a trade agent in which an analyst who knew little about the process worked with a senior specialist who knew it thoroughly.

 

Bhattacharya said she looks for two things when hiring: the ability to learn and hunger. Competencies, she said, can be picked up quickly today.

 

 

 

Trust, bias and cyber risk

Asked what keeps her awake, Tan said cybersecurity.

 

"This is not a competition anymore between banks," she said, because an attack could hit telecoms, utilities and other critical institutions alike.

 

Defences include using AI to find AI-driven threats, basic cyber hygiene, and multiple layers that make attacks harder. Because data is shared with partners, from telecoms firms to printers, she said her "Project Clear" aims to give every piece of data a clear owner and accountable party from creation to deletion.

 

Agents must also operate under the right governance, with a control plane that evaluates what they do, checks they stay within their remit and keeps privileged access safe.

 

On the risk of AI agents telling users what they want to hear, Bhattacharya said Salesforce routes AI use through a "trust layer" that encrypts and masks data, ensures models do not retain it, and screens for bias, toxicity and hallucinations.

 

Hallucinations have been reduced but not eliminated, she said. In consumer-facing chat, where systems try to reassure users, "you need to keep your human brain on".

 

 

 

What stays human

Makiko Ono, chief people and culture officer of Suntory Holdings, said people are the group's most important asset. Because Suntory sells consumer goods intended to enrich people's lives, she said, 'AI unlocks possibilities, but people create the value.'

 

She cited whisky, which matures over long periods.

 

"AI cannot make whisky," she said, and the same long-term view applies to staff, who also mature over time and cannot be judged quickly.

 

Suntory uses AI for the first screening of recruits, followed by human interviews, because it cannot be trusted completely, she said.

 

It also runs a reverse mentoring scheme in which younger staff in the digital department tutor senior leaders in using AI, which she said is working well.

 

 

Young workers: easier and harder

Asked whether young women now have it easier, Bhattacharya said it is easier in some ways, because knowledge is democratised, and harder in others, because competition is greater and new roles such as agent orchestrator and agent evaluator are emerging.

 

The issue is not specific to women, she said, but a wider social one, causing dissonance among young people. Older generations, she said, must give them hope and purpose, ensure AI is used fairly and help them find meaningful roles.

 

Closing the session, one panellist offered young people three words to hold on to: curiosity, creativity and courage.