DBS Bank's CEO Piyush Gupta On Using Technology To Police Technology

NOTE: This transcript contains the host's monologue and includes interview transcripts by a machine. Human eyes have gone through the script but there might still be errors in some of the text, so please refer to the audio in case you need to clarify any part. If you want to get in touch regarding any feedback, you can drop us a message on [email protected].

Piyush, thank you so much for joining me. So, we are going to go on a little journey and your digital journey and DBS's digital journey in the last decade or so. So, tell us about where you started thinking about technology in a very applied sense and then we'll talk about how it fits into the context of artificial intelligence and what are the kind of problems it's going to solve.

Well, you know, we actually we're going to start with my own background. So, I used to work with Citibank and in the late 90s, I started fooling around with a lot of the first e-commerce platforms in Asia. In 2000, I quit banking for a year and set up my own dot-com in Delhi actually with the Hindustan Times people.

So, I had some affinity and some,

What was it called?

Go4i.com. So, we sold it. HD online today is what we created at that time. It didn't work by the way, dot-bomb rather than dot-com. But, so I had some background.

In 2012, two or three things happened. We've been trying to grow our business around the region through the traditional brick and mortar ways and I found that SME and consumer banking were very hard to scale. I also found that the financial services value chain was beginning to get unbundled around the world.

But, perhaps most important, I had a seminal meeting with Jack Ma at Alibaba and it caused me to think that these guys were redefining how financial services would operate. They had 3 million SME customers lending, they had 300 million consumer customers, you know, again, borrowing customers. They were raising money through Eurobow, they were moving money around, they were selling insurance and they had zero branches, zero salespeople, zero paper.

So, you know, they were a bank to all intents and purposes. So, that caused me and my CEO to sit back and say, hey, you know, if the tech companies can do it, it will happen. And why could we not do it?

You know, technology is open source, it's available. We have domain knowledge, so we could leapfrog. We were somewhat lucky that we hadn't got impacted by the global financial crisis like a lot of the Western banks.

And because we hadn't, we could focus on the future instead of focussing on the past. So, 2013-2014, we were very early in the transformation game. We set up a programme of change, which was architected around three pillars.

The first was the technology re-architecture. So, very early, we moved to microservices and an API-centric technology. So, we set ourselves four years to really completely change the nature of our infrastructure, move from the mainframes to x86 server farms, etc., etc. We created our own virtual private cloud, and so on to the tech pillar. The second pillar, which is actually quite germane to this year's NASSCOM, we figured that it was finally about the customer. I mean, you had to be customer-obsessed, not just customer-centric, but customer-obsessed if you really wanted technology to make a difference.

Like I told my people at that time, they would argue that Uber became Uber not because they used technology, but because they reimagined the job to be done. Airbnb is not like Ritz-Carlton didn't have technology, but Airbnb reimagined the job to be done. So, you really had to be customer-obsessed and think about the human agenda behind technology, if you could.

And so, we really doubled down on this whole idea of customer journey, customer thinking, outside in to leverage technology in a very different way. And the third pillar was the culture change. You know, we figured if you wanted to compete with the Googles and Amazons and Alibabas of the world, you had to believe you're a tech company.

You could not. And to believe you're a tech company, you had to take everybody in the journey. So, that was the other agenda around being human-centric.

We had 18,000 people at that time, and we sort of called ourselves the 18,000 people startup. You know, 40,000, we still call ourselves the 40,000 people startup, because you have to make sure that your people feel part of the change. There cannot be a set of people who think change is happening to them.

They must believe they're driving the change. Then you get really superior outcomes, right? So, we started the journey then.

We started getting, you know, fortunate. In 2016, we were recognised as the world's best digital bank. And since then, we've been, you know, recognised that for several years in a row.

So, you know, you talk about Alibaba, and you talked about the fact that, you know, they have so many millions or hundreds of millions of customers, they have vendors, and there's money flowing back and forth. But at the end of the day, they are an e-commerce company.

Now, you are a bank, your job is also to keep deposits of your customers, large and small, safe, and answer to regulators. So, how do these two worlds collide at all in the ambition to be more like a tech company versus the traditional responsibility?

That's a really good question. Frankly, if you ask me, when I look back, what is the biggest challenge been? It has been that.

How do you get the balance right between the fiduciary role you play as a bank and the regulated nature of our industry, but on the other hand, trying to be a tech company and be in the context of the customer? I will start with a fundamental principle that, see, nobody wakes up in the morning saying, I want to go banking, you know. So, people want to do different things with their life.

You don't need a mortgage, you want to buy a house, you don't need a credit card, you want to pay for your meal. And so, we start with that first principle that people actually want to do different things with their life. And then you figure out how do you, you know, embed your banking products in what people want to do in the context of what they want to do.

You start thinking like an e-commerce company. I, you know, very early, we started saying our role is to make DBS invisible. Embeda says, and today it's called Embedded Finance.

So, this was way before it was called Embedded Finance. I wanted to be the Intel inside of banking, right? And so, you go and put this thing out there.

And in working with that, there were two or three things, the balance on other side. One was the regulatory environment. So, what would regulators let you do?

But the second was the trust environment. So, how do you actually make sure that you don't get to a state where customers stop trusting either because of fraud, scam, cyber security and crime? So, you know, and sometimes they coincide, the regulation and the trust environment.

On a regulatory front, we were fortunate that regulators by and large in Asia have been willing to lean in. They've been progressive and they're willing to actually, you know, work with the banks to find constructive solutions that allow the industry to progress. And we're fortunate because when you compare the regulatory response in our part of the world to the West, quite different.

You know, if you go and look at the Fed or even the PRA in the UK, they were still coming from the aftermath of the global financial crisis. And for the bulk of the decade, they were drawing hard lines. Whereas regulators in Asia are willing to work with sandboxes.

They're willing to work with experimentation to say, how do you get the best out of the technologies and not, well, not throwing the baby out of the bathwater. And so, we got fortunate that many of the experiments we did were actually with the regulators. In Singapore, where we're headquartered, the MAS, Monetary Authority of Singapore, created something called a Financial Centre Advisory Panel in 2015-16, where they got the e-payment companies, the banks, the regulators together.

And we get together every six months to think about how we could leverage technology differently. So, that was actually quite helpful. But the bigger challenge was this issue around trust.

And trust in many dimensions. The first is just resiliency. You know, so the customer's expectations are now driven by Google or by Amazon.

They want 24-7. They want instant. And therefore, you need a very high degree of resiliency.

And actually, this is one of our lessons. We didn't get that right. And so, you know, we ran into a state where we moved to the new architecture without recognising the operational complexity it involved.

And so, we had to go back and, you know, course-correct and so on. So, you need resiliency. But on top of that, you need to be very concerned about cybersecurity, right?

So, in the old days, if somebody frauded a customer, it was one customer at a time. Today, if somebody gets in, it's a million customers at a time. So, thinking very hard about cybersecurity, peripheral defences, firewalls, and then increasingly, you know, we have a mantra which says somebody's going to get in.

And therefore, I call it inside is the new outside. You've got to assume somebody's inside. And then you've got to figure out how do you catch the, you know, perpetrators.

So, using AI, micro-segmentation, data, this thing, circumstantialization, use all kinds of technologies and techniques to create customer trust. But on top of that is this whole new world with scams and frauds. I mean, people are constantly…

And now with AI, deepfakes. So, we're constantly trying to figure out how do you use technology to police technology. You can't do it by human beings alone.

And it's a constant cat and mouse game, you know. But if you don't, you know, programme for that, you will lose the battle.

So, I'm going to come to that deepfake and those challenges in a bit. But tell us about how you look at consumers in terms of the way you're trying to understand them today. And since we're in a technology conversation, using, let's say, the kind of AI that is available. And to what extent does it even help you do that? And the second sort of supplemental question, if I can add to that, you work in 19 countries. So, are customers different, completely different, somewhat similar? How do you mine the similarities and then work on the dissimilarities as you do this?

All right. So, first of all, you know, this whole idea of embedded finance and being able to predict what a customer really wants, AI is a massive game changer. So, we got on the AI journey about a decade ago, more than that.

We were the global pilot for IBM when they came up with Watson. And so, we were the global pilot for wealth management. And didn't work that well at that stage.

I was getting accuracy rates of 85, 87 percent, which is not good enough to take to market. NLP hadn't advanced to the level that you'd like it to be. Unstructured, you know, PDFs, graphs, you couldn't read that kind of stuff.

But fast forward over the next few years, we learned. And 2017, we started scaling up predictive AI in a big way. We created our data lays, created a whole bunch of data, access protocols, etc.

We started creating models. And we found that we were beginning to get really superior outcomes in using AI to be able to predict and to be able to help customers in the servicing needs as well as the sales process. So, let me give examples.

You know, a human being, we've got like 20 million plus minus customers. I have X number of RMS. So, by and large, you can handhold and deal on a relationship basis with a handful of customers. The rest of them are, you know, faceless.

But with AI, they are no longer faceless. You can start dealing with a customer, a segment of one, in a relationship way. The amount we actually know about customers is massive.

I have unbelievable amounts of data, you know, what you do. I can tell you, for example, the high degree of accuracy, what restaurant you're likely to go to in the next month, because the data can tell you that. It shows me what you like.

But individually, I can't do it. So, AI can help a lot. So, we started building a lot of AI models.

Today, we have 1600 models driving the bank, 400 different use cases. We send out 50 million nudges in Singapore alone, every month. And those nudges help you figure, hey, this might be a duplicate payment.

Hey, you're paying too much for electricity bill. Or by the way, you were reading about gold, and this is a good time to go and maybe buy gold. And this is all powered by AI engines.

We're finding engagement rates are up. Satisfaction rates are up. Our response to our marketing nudges are up.

Three, four years ago, I started doing A-B testing to capture the economic impact. We run an A-sample without AI and then the rest with AI. This year, we'll make over a billion dollars of economic delta by use of AI and AI models.

So, without doubt, it's been a game changer in terms of our ability to predict and be able to facilitate customers in terms of what they need. And that's frankly even before Gen AI. I think Gen AI turbocharges it to a different level.

So, you're also saying, therefore, that customers have a great degree of similarity. I guess we know that, but from your vantage point.

That was the second part of your question. So, you know, it's interesting. I think our 19 countries are mostly in Asia.

I find there's a lot of cultural similarity in Asia. The big difference between Western liberal thought and Asian thought is the primacy of the individual and data primacy relative to the role of community and society. And I'll tell you an interesting anecdote.

When COVID happened, in the first day, we had the first COVID case in February, my data analytics and people came and did the usual networking analysis. And we were able to tell using, you know, the door tap data and QR codes, you know, the people who might have been in touch with this person. And our employees loved it.

I did an op-ed in the Financial Times saying how we'd use data and analytics to be able to decipher who might be at risk. I got flamed in the Western press because for them, the notion that I would use employee data was unconscionable, whereas all my employees in Asia loved the idea because it was a community service. So, that difference is a real difference.

But in Asia, I find, whether it's the Chinese or the Indians or Southeast Asians, the sense that you can use data constructively with safeguards and you can do things is actually quite common. And so, we've been able to find a lot of ability to success transfer across most of our countries.

So, as you apply AI now to, let's say, real world consumer needs, perceived needs, because you're obviously trying to pull them out even before they think of them, what are some of the interesting use cases and how is it scaling up? And then I'll come to India as well.

Well, you know, if you go back to general AI, I said, see, you take the mortgage journey, right? So, typically, people come to us to take a home loan well after they've chosen the house, well after they've decided, and then they shop for the mortgage rate. Whereas if you do embedded finance, you can start engaging with the customer six months ahead of time.

Why? Because your AI, you know, robots and engines can start figuring and sensing this customer might be interested in looking for a house to all the other data it puts together. So, you start engaging with the customer much earlier.

And then you also start figuring from AI engines what the customer is likely to respond to. Is it likely to be a service that helps them find the house? Is it likely to be a better mortgage product?

AI engines determine all that and decide when to approach the customer, what kind of offer to give the customer, how to engage the customer. It changes our customer engagement relationship quite considerably, as an example. As you go forward today, we're finding that with gen AI, because you can deal with unstructured data, anything I read, synthesise, you know, create output, the computer does much better than individuals do.

We're already finding the impact of that to be really helpful. So, our call centres where customers used to call us, we had chatbots for some years. But now with gen AI, it's just a completely different level of engagement with the customer.

The customer is getting more and more comfortable asking the questions. We have an enterprise knowledge base in the back. The gen AI tool goes and engages with the customer much better than our agents, physical agents can.

So again, the customer satisfaction engagement levels are improving quite dramatically. We use it for other stuff as well internally. We're using AI for hiring.

We have something called Jobs Intelligence Maestro, which actually predicts what kind of sales people or what kind of computer people. So, we use it for hiring staff. We use AI for onboarding customers.

Our underwriting process is no longer based just on financial statements and cash flows. We predict what kind of customers we could bring on. And so, we're now able to reach customers who we would normally not be able to reach.

So, all of these are use cases that we're being able to leverage.

So, is AI then to that extent resolving business challenges or improving your business outcomes better than perhaps what was this thing or is it too expectation or it's there's some gap?

No, no. So, it's without a doubt, resolving challenges much better, but there is still a long way to go. So, when I look even at the predictive AI, I tell people, if you look at the scale of 10, we think related to the banking universe, we are top quartile. But compared to what a Facebook does or Netflix does, we are in three and these guys are at eight or nine.

So, there's a lot more we can do. With gen AI, the real trick is going to be to reimagine business models. So far, we are getting productivity within existing business models, but you're going to have to rethink the way work is organised.

You're going to have to rethink organisation structures. You're going to have to rethink how you go to market. Gen AI gives you tremendous capability to outsource to the customer.

So, you don't even have to do the work. The customer will do the work for you, but it requires you to put on a fresh hat and say, how could this be completely different? So, I think we made progress.

It's much better than it would have been without AI, but relative to where it can go, we still have a lot of work.

And when you say outsource to the customer, is it like how I'm printing my boarding pass at home and then going to the airport?

It's sort of like that, but one step beyond. In the future, I can see an AI model bank. So, every customer has their own bank and the bank itself is built and driven around AI and you basically deal with your own bank in your own way, right? You can get to the stage where, you know, a large part of what we do could be renowned.

This could take three, five, seven years, but you can get there. So, where do you get these thoughts from? As in, when you think of something like this, what's the inspiration?

Well, it's a lot of it is, you know, reading and talking to people. I talk to people like yourself and read a lot, but more than that, we have two things. One is, we have a very structured innovation process, which we institutionalised some years ago.

And so, we innovate at four levels. Obviously, there's the bottom of the pyramid, you know, the old suggestion box, except it's all digitised. And then we have something, we run the bank through journeys.

So, we have 63 journeys and we don't run the bank vertically anymore, we run horizontally. Every journey and lead and sell is charged, it's like agile at scale, but every journey lead is charged with coming up with innovations that make their journey superior and better. So, they come up with ideas.

Then we have some tech platform, let's say payments is a platform. That's the third level. They're charged with imagining the future of payments for the next three, five years.

So, they'll come up with their own thinking. But what's really important at the top of the house, we meet every six months. And there, based on all of the research and so on, we make one or two big bets on where we think the future of the world is.

So, we made a big bet on AI some years ago. Three, four years ago, we made a big bet on tokenization and the digital token economy. So, we have one of the first regulated entities that created our own crypto exchange as an example.

So, we use all of this to try and, you know, programme for some of these ideas.

What's the latest bet?

Actually, we're still doubling down on Gen AI. I think the scope is huge. And similarly on, you know, crypto and the digital economy, it's a good start, a good business.

We made 30, 40 million bucks on it. I think that could be much, much larger. So, we're just currently focused on these two areas to drive a scale and scope.

And crypto, of course, not in India and regulators look at it a little differently here.

But what you got to think about, it's not, you know, crypto is one part of the use case. The real thing to think about is blockchain and digital assets. So, I think everything is going to be digitised.

Yeah. And the real power of a blockchain digital asset is you can programme. So, in Singapore, for example, with the regulator, we are running something called programmable money.

And so, when you actually do, it's a smart contract. The contract embeds the programme and that only comes into effect if certain conditions happen. That changes the back of a structure completely without cryptocurrency.

Tell us a little bit about India. What are you thinking of here? And are you happy with the way things have progressed?

Well, we are actually finally quite pleased. You know, we tried to do some interesting thing. In 2016, we launched our first digital-only bank in India. It was intended to be only on a mobile phone.

And we learned some, you know, hard lessons in the first two years. First, like everybody, we put on 3 million customers in the first year. But when you looked at it, you realised that these are eyeball kind of customers.

You could not monetise these customers in 30 years, right? The people were the young kids who were basically gaming for the freebies. So, we had to course correct a little bit.

We also realised something about the infrastructure at that time. So, pure mobile doesn't work because you keep losing signal. So, we had to add a desktop internet capability to supplement the mobile.

But even after that, we realised that for a country like India, you probably need fidgeting and not just digital. You still need trust and you need brand relevance. So, in the middle of COVID, we acquired this bank called Lakshmi Vilas Bank.

And so now we have a mix of both. We have 500 branches in 350 cities coupled with our digital capabilities. And now we're getting some really great outcomes.

It's the fastest growing business anywhere in all our 19 countries. So, very, very positive about the India situation. We used to be a large corporate bank, but now we're getting scale and scope in both the consumer business and the SME business, if you will.

Obviously, the digital infrastructure in India is superior. It helps. The Aadhaar stack, it helps.

I mean, the India stack helps. The e-commerce agenda helps. The payment rails help.

So, we've been able to leverage that very effectively to turbocharge our growth.

And any specific India objective or plan for the next year or so?

Well, there's no objective or plan for the year or so, but I mean, if you had to step back and say, what is our big picture? We'd like to be an embedded local bank in India. So, the difference between us and most other foreign banks is we are trying to be another Kotak or another HDFC.

So, our ambitions are very different from the typical foreign bank ambitions. And we are happy to play the long game to get there. Right.

That's a good note to end on.

Thank you so much for speaking with me.

All right. Take care.

Happy to do this. Thank you.