Banks should be prepared for providing credit through UPI: Zeta CEO APAC

We believe that the UPI transaction volume, using one or another form of credit, would be $1 trillion by the end of 2030, she says

Bengaluru-based Zeta, which provides banking technology to institutions globally, is looking for opportunities in India. It serves HDFC Bank in India and aims to strengthen the core banking systems of other lenders. RAMKI GADDIPATI, chief executive officer (CEO) for Asia Pacific (APAC) and global chief technology officer (CTO) of Zeta, in a video interview with Ajinkya Kawale spoke about technology and banking. Edited excerpts:

How has the first half of 2024 been for Zeta?

It is an exciting year for Zeta. We will be doing some work in the US. This year, we have decided that we’ll start our business development in India as well, with our digital credit as a service as an offering for Indian customers. We launched this in May, and we conducted an event inviting all banks in the country to familiarise themselves with UPI (Unified Payments Interface) credit, and how that can transform the credit landscape in the country.

Which products will see demand increasing in India? 

Within India, for the last three years, we have been exclusively working with HDFC Bank. We worked with them to relaunch their PayZapp, which has over 11 million users. We also launched a credit card product with HDFC called Pixel, which is the country’s first digital native credit card. HDFC Bank is probably India’s first Cloud native core banking deployment. 

We demonstrated that we could handle 200 million credit card accounts and do about 10,000 concurrent transactions, and process what is considered as extremely hard in banking: The end-of-day account processing for all these accounts. 

Is there a lot of catching up to do with technology in Indian banking?

I think we are preparing for the future. Digital transactions led by UPI have grown in India, and we are looking at an upwards of a 30 per cent CAGR (compound annual growth rate) on digital transactions. The UPI growth has happened only on the back of most savings accounts and current accounts of users. 

We want to make sure that credit can grow on a scale similar to digital transactions, using rails similar to UPI. The National Payments Corporation of India (NPCI) has enabled it, but banks should also be ready and there are notable gaps here to make a change happen for credit. A lot of core banking systems can scale and meet the demands of UPI today. 

If a similar volume of savings account-based UPI transactions were to move to credit products, banks may inherently never be able to cater to it. A credit line on UPI is going to make banks think about credit fundamentally as they digitise the credit process entirely. We will be in a place to be able to enable banks to do that.

Banks face outages and sometimes  transactions fail. What is the tech gap at banks? 

At the outset, every bank wants to solve this problem, and the challenge is well recognised. However, these problems are not easy to solve as migrating a core banking system is not something that many banks can afford. It is extremely complex, and in case one wants to migrate what alternatives do they have? Very few banks are able to make that change because of the challenges and the costs involved. Other banks may be looking out for a staggered approach to the solution. 

For instance, in India, we are saying that we will work with banks to create new credit products so that they themselves do not face similar challenges. We are telling them that not to fall into the trap again with credit products. It has hurt banks for the savings product, and they may lose dollars for a credit product. 

How big is the credit line on UPI opportunity?

We believe that the UPI transaction volume, using one or another form of credit, would be $1 trillion by the end of 2030. It is not going to be linear growth, but a hockey-stick growth instead. It is the preparation of the ecosystem towards it that is going to take a few years. It's not just banks but also merchants, vendors, among other stakeholders since in the present set-up UPI transactions are free to merchants, but later on, credit cannot be free. There is also some ecosystem level education that needs to be factored into the ecosystem. 

The interchange fee for credit lines on UPI has not been defined yet by the NPCI. Fintechs too are going slow on retail credit

What we are talking about is not for the next two months, or a few quarters. It is an infrastructure-building exercise at the bank’s end. Typically, based on our conversations with banks, it takes six months to one year to even decide on the entire system stack that they would put to use services. It is not a trivial decision. Moreover, the important thing about credit is that it is not a blitz-scaling story. 

To extend credit responsibly, you need to know the customer using that credit. It is not just the credit score to be taken into account, but you need to build a relationship with the user to understand their behavior and gradually grow that relationship to build a credit pile that is worthy of whatever the value one will be commanding. It is not about knowing the algorithms as much as actually knowing the cohorts of users that one will be dealing with. It is not just a data problem, but also a human problem.

What does a system stack include from purely a tech perspective?

We have a benchmark that every Application Programming Interface (API) should respond under 200 milliseconds.

There is a strong customer identification and anti-money laundering (AML) stack. These are core functions in the process that banks and regulators call CDD or customer due diligence. For underwriting credit products, one looks at credit risk, identity risk, a person’s political exposure, among other things.

The second set of systems with respect to credit risk underwriting, continuous monitoring and limit management is called the risk limit management. To give credit, there is an origination journey, then provisioning loan management systems.

Then you have various disbursement systems, the actual transaction, ecosystem transaction switches, payments, issuer processing systems for processing transactions, and a fraud and risk management system. There are also Clearance and Settlement Systems, which become what is called a transaction processing ecosystem.

Collections can be broken down into repayments and collection processes. Repayments happen over instruments like UPI, net banking, debit cards, among others. Collection systems remind customers, nudge them to make payments, informing them of consequences of not having a healthy credit score. There is analytics, machine learning involved in customer interactions that are managed by the collection systems. 

How does Zeta compete against well-entrenched players who are already offering tech services to banks?

There are barely any players who have the kind of breadth and depth of stack that we have in the space that we operate in. We deal with some of the largest banks across India and the US, with a completely cloud native stack.

We have invested more than $300 million into building this stack, and we continue to invest upwards of $40 million every year in the research and development (RnD) to further evolve this product. The extent of investment that we have made into this is not going to be matched by any player any time soon. Since software cannot be seen, the comprehensiveness of the depth of such products is not felt.

How do you view the rise of artificial intelligence (AI) and generative artificial intelligence (GenAI), including in the financial services space?

I remember 7-8 years ago, people used to ask what is a mobile strategy, or a banking strategy. Today, all of that is deemed as natural expectations and similarly AI is going to be commonplace across the board in everything one does. There is going to be a lot of productified-AI.

Unfortunately, for the banks, for the current product that they have, can AI be integrated into the majority of their products in the next four or five years?The answer is no. While they all recognise it is necessary, can they become cloud friendly? No. Similarly can they become AI friendly? Not so quickly.

But, despite AI kicking in, would you still continue to see demand for workforce over the next decade even as automation becomes a commonplace?

In the product stacks that we have, with all the automation that's in place, one would require less people than those required operating a similar product using a legacy stack. That being said, in the short term with the competencies and skills required to operate the new way of thinking and new model of products, there will be employment creation for such skills.