Exclusive: DBS bets on AI to scale customer-centric banking

DBS Bank began its digital transformation journey long before generative AI became a buzzword. Over a decade later, the bank says its efforts have not only modernised its infrastructure but are also delivering hundreds of millions in value – with sights now set on the billion mark.

"Banking used to be one big boulder that you presented to a customer," said Nimish Panchmatia, Chief Data & Transformation Officer at DBS Bank. "But fintechs started coming in and unbundling it. We could sit and watch, or we could do something about it."

The bank chose the latter. The emergence of new players in payments, lending and buy-now-pay-later models signalled that customers were demanding more flexibility. At the same time, technology companies like Google and Netflix were setting a new benchmark for speed and experimentation.

"It used to take us 18 months to deploy a solution. These guys were doing 1,000 tests a minute," he said.

DBS decided to reinvent itself. The result was a full-scale transformation programme internally dubbed "Gandalf" – an acronym for Google, Amazon, Netflix, Apple, LinkedIn and Facebook. The mission was clear: become as tech-savvy as the companies they studied, not just as good as any other bank.

"We were going to be the 'D' in Gandalf," Panchmatia explained, "meaning DBS would be on par with these global tech giants."

The transformation focused on three pillars: reengineering technology infrastructure, improving customer journeys, and reshaping internal culture. The bank migrated from legacy mainframes to microservices and open-source platforms, all while building a virtual private cloud when regulators were still wary of full cloud adoption.

"The regulators told us, 'do anything you want, just not the cloud'. But we saw the power of virtualisation, so we created our own," he said.

Reimagining customer journeys meant working backwards – not starting with bank products, but with what customers actually wanted. "We did customer immersions," Panchmatia said. "We asked them, 'if this wasn't banking, what would your expectations be?' The answers were very different."

Perhaps the most radical change came from within. "At the time, we were 22,000 people," he said. "We wanted a startup culture where it was okay to fail – not burn the building down, but try things. And if it didn't work, what we learned was more important than doing nothing."

This internal mindset shift enabled experimentation with data analytics and machine learning as early as 2015. One standout success was optimising ATM maintenance in Singapore, which has some of the world's busiest machines.

"We applied predictive analytics to determine when parts would wear down or when cash would run out," he said. "At our peak, machines ran out maybe four or five times a month – compared to hundreds a day before."

Customer satisfaction soared, and the bank began to believe it could use data to solve much bigger problems. But as more teams embraced analytics, a new challenge emerged: too many different datasets, each interpreted in its own way.

"You can't have two different versions of the facts," said Panchmatia.

That triggered the next phase – consolidating data into a central lake with a single platform, unified governance and a clear ethical framework. The result was a system named ADA (Advancing DBS with AI), surrounded by a protocol called ALAN, and guided by principles dubbed PURE: Purposeful, Unsurprising, Respectful, and Explainable.

"We asked not just 'can we use the data?' but 'should we use it?' That's a very important question," he added.

The bank also focused on training staff to feel comfortable with data, offering visualisation tools and basic scripting capabilities, while creating partnerships with universities to develop local analytics talent.

By the time generative AI arrived, DBS was ready. "We had the culture, the tools, the people. We were familiar with the work already," Panchmatia said.

Today, generative AI is available to every DBS employee. Staff can rewrite documents, summarise content, or query the organisation's knowledge base. "You can walk into a branch in Bali and ask about opening an account in Singapore, and the person will be able to help without making a phone call," he said.

The bank's work in AI is already delivering measurable value. "In 2024, we were able to attribute $750 million Singapore dollars of value from AI. This year, we're gunning for over a billion," Panchmatia revealed. "We're one of the only banks in the world that can attribute AI value in a clear way."

That's not just in back-office operations – AI is helping advisors prepare for bespoke meetings with both private and corporate clients, integrating data from research, CIO reports and more.

So, why did DBS partner with Google?

"The biggest reason was their customer centricity," he said. "From solution design to the CEO, there was engagement and commitment. We may not have purchasing power, but we've got brand power – and Google respected that."

Most of the bank's generative AI workloads are now running on Google Cloud, with India recently going live. "It was done in record time," Panchmatia said. "It's a great spot for them and a proud moment for us."

DBS now runs roughly 30-40% of its compute in analytics, and Google Cloud forms the beachhead of that strategy. Indonesia is next on the roadmap, with other markets likely to follow.

Looking forward, Panchmatia sees generative AI as a game changer in customer engagement.

"It gives us far more opportunities to engage with our customers than we had before," he said. "We'll use this opportunity to make even more wonderful solutions for our customers."