HDFC Bank, India’s largest private-sector lender by market value, has been deploying digital assistants and robots in branches and using Facebook Messenger-based bots to deliver new, interactive experiences for its customers. In a conversation with TechCircle, Nitin Chugh, country head of digital at HDFC Bank, explains how the lender is using emerging technologies such as artificial intelligence (AI) and machine learning (ML) to improve customer experience and manage operational processes better. Excerpts:
What has been your approach towards emerging technologies?
We are looking at all the emerging technologies such as AI, ML, internet of things and blockchain to harness them for providing better services and streamlining operations.
We are mostly using AI, ML and blockchain. There are majorly three focus areas when it comes to the emerging technologies. These are giving customers personalised and intuitive experiences via new interfaces, adopt strategies based on digital analytics and streamlining various operations.
To what extent are you embracing these technologies?
We have been working on AI for two years now and trying to divide the whole work into three or four layers. For example, one layer is for customers to have a digital experience and the second is to improve bank interaction. The third is including AI in our own internal management processes and the fourth to run processes in non-core operational areas.
We are using AI to remarkably change the experience for customers. This can be divided into three or more categories to include our chatbots, our digital assistant EVA and our humanoid IRA. The entire effort has been to make interactions more conversational.
We have developed two chatbots with help from startups such as Niki.ai and Senseforth Technologies to help customers have an easier communication channel with the bank.
The EVA–now available on Amazon Alexa on Echo devices and Google Assistant on Google Home and Android smartphones–was also developed to answer intelligent questions for customers. The IRA, which now has a sister at one of our Bengaluru branches, provides a unique interaction platform with customers.
We are trying to figure out a way to use AI to develop personal experiences. We are also using AI to run internal processes such as risk management and portfolio management. Lastly, we are using AI and ML in non-core function areas such as hiring. For example, the use of AI helps us screen candidates and conduct psychometric tests.
In short, we are trying to adopt the full practice of AI in such a way that it doesn’t remain just a service but becomes one of the fundamental stacks in our tech architecture.
What does it mean for the customers and the bank?
We are looking to provide more personalised experiences to customers. Customer experiences can become impersonal with digital, which involves interacting with a mobile app or a web browser. But AI has the ability to provide a more personalised experience.
One can use natural language processing (NLP), and over time introduce deep learning on voice, among other things, to drive these experiences to a superior level. The idea is to take it to a level where these interactions become better than face-to-face interactions.
For the bank, there are multiple use cases. For example, the chatbots save a lot of time and have been in demand as some of our customer segments are already open to messaging. We also have a chatbot-like digital assistant for our knowledge bank. The chatbot helps in finding circulars kept in the knowledge bank and show it to customers.
Could you explain the use of AI and ML in contextual marketing?
AI and ML in contextual marketing is a separate project and is part of our analytics practice. What we are trying to do is to use AI in such a way that we get answers to questions such as how to change marketing efforts. For example, changing a 30-day offer to an offer valid for 30 minutes. While it sounds simple, there are a lot of components into such a service.
The first thing that you need is the ability to detect the signals such as user behavior or pattern. The next step includes making some sense of the signals received and this is where AI steps in. Once the AI has run its course, you should be able to generate some actionable intelligence that finally is moved towards the customer.
Let me explain with examples. We are testing a new mobile app for the bank. In the old app scenario, if you wanted to make a bill payment, you would find an SMS saying that a bill was due. You would log into the mobile app and then carry on with the transaction.
In the new app, what happens is that the application will pick up a signal saying that the bill is due, as soon as a new signal is generated that the customer has logged into the phone, instead of hoping that the customer will remember about the bill. The notification for the due bill will bubble up and take the customer straight to the transaction on the screen.
There is another test case that we have developed but not rolled out yet. If you want to hotlist your credit or debit card, you can start a conversation with the bot in the app to block your card. We are also bringing a feature that will start pushing offers and discounts based on a customer's location. As soon as the customer changes location, the context will change and the customer will get a new offer.
Can there be other use cases using the same logic to help customers?
The first thing that comes to my mind is the security aspect. About a year ago, we developed a security feature to avoid misuse or fraudulent use of debit and credit cards.
Along with National Payments Corporation of India and a startup, we were able to map the location of a customer through the phone not with the help of GPS but via triangulation of telecom towers and then match it with the terminal ID of the ATM that the customer used.
Invariably, the assumption is that 99% customers would carry their phone when they use an ATM. Now if the phone and the ATM are at different locations, there is a high chance someone else is using the card. In that case we send you an alert to block the transaction.
How do you use AI and ML for portfolio management?
Portfolio management is not only about managing risks from a credit point of view but also includes day-to-day stuff. We have been using AI to find out how you want to segregate your customer base, study their behaviour, come up with the next set of actions and design your marketing campaign.
Earlier, you had to run queries or macros to get insights into the data. But today, with AI you can get faster data analysis with better insights to make decisions.
Are you using AI to analyse the credit profile of customers?
To do credit profile analysis, you need loads of data. While we have our own data sets, we have partnered with startups such as CreditVidya and a few others to gather secondary data that would lead us to ancillary markups such as social scores. Once we have all the data, our pre-programmed ML algorithms run through them and come out with HDFC Bank score that suggests if we should go ahead with the risk.
A consortium of banks is partnering with companies such as IBM to deploy a blockchain network for information sharing. Is HDFC Bank involved? How is the bank approaching blockchain?
Although we are a part of the consortium, I cannot comment on the blockchain projects that are in the development phase. I can only tell you that though the country has taken over two years to understand the new technology, you can see some rapid deployments of the technology this year.
HDFC Bank has also been experimenting with the technology and has partnered with several companies to develop proof-of-concepts for various cases. However, we cannot divulge more details at this time.