Using AI/ML to cut cost-to-income ratio, making fintech bets: ICICI Bank's Madhivanan

Emerging technologies such as robotic process automation (RPA), artificial intelligence (AI) and blockchain have changed the way financial companies conduct business, and ICICI Bank is no exception. In a conversation with TechCircle, B Madhivanan, chief technology and digital officer at ICICI Bank, said the lender is using newer technologies as part of a strategy to bring down the cost-to-income ratio. He also talks about the bank's fintech investments. Edited excerpts:

Newer technologies have had a serious impact on banking processes. How is ICICI Bank using such technologies to streamline operations and smoothen customer experiences?

ICICI Bank was one of the first banks to start using emerging technologies but we did it in phases as technology keeps evolving at a fast pace. Our first customer experience point was the ATM post, which we moved into core banking services - a product we co-developed with Infosys - offering customers more agility in banking. Then we extended into retail banking and corporate banking. 

In the second phase, between 2013 and 2015, when iPads and Samsung tablets became popular, we used it to help our sales team acquire customers or aid our employees in wealth management and other such activities where the tech hardware and software assisted our staff to get work done faster and more efficiently. 

The third phase was more interesting because as the number of customers grew for the bank, we started looking at automating a few processes and later we moved to AI. We are continuing to use AI and machine learning (ML) to streamline more of our operations. We were also one of the earliest banks to complete a blockchain process.

You talked about automating processes in the third phase. Could you explain?

I was talking about robotic process automation (RPAs). We introduced RPAs, calling them 'software robotics', as early as September 2016 in over 200 business processes. These automated processes emulated human actions to automate and perform repetitive, high-volume and time-consuming business tasks cutting across multiple applications. 

They reduced the response time to customers by up to 60% and increased accuracy to 100%, thereby sharply improving the bank’s productivity and efficiency. This enabled the bank’s employees to focus more on value-added and customer-related functions. 

The software robots then performed over 10 lakh banking transactions every working day. Our retail banking was growing at over 25% annually during that time. Currently, we have automated about 950 processes and soon we intend to include RPA in at least 1,500 of our operation processes. 

The customer has no need to visit branches anymore. More than 82% of our transactions happen digitally, followed by ATM transactions. Branches only account for 4.5% of transactions. Branches are now being opened only at a case-by-case basis.

When and for what processes did you foray into AI?

There are many processes or activities that the bank conducts and we rank them on the level of complexities involved in the IT department. The simple tasks are tagged as L1, for which we use RPA. The second level of tasks, called L2, involves a little more complexity and requires calling more than three to five activities or functions. The most complex tasks are tagged L3. It is in the third level where we are making the use of AI and ML. 

We work with more than 14 to 15 fintech companies in developing these solutions and a few examples would be CreditVidya and Niki.ai. In fact, we run innovation labs based out of Mumbai and Hyderabad and have made investments into four of these companies.

But there is another side to the AI story. AI can tremendously powerful in customer satisfaction and answering their queries. We launched our first AI-led solution called iPal, a chatbot, in February 2017. This is available on the iMobile app as well as the bank's website. The bot has natural language processing abilities and can talk to customers in English. We saw huge demand from millennials who would rather chat with a bot instead of calling up the customer care. The AI-bot can solve more than 75% of simple queries and tasks for customers with more than 90% accuracy.

How about integrating iPal's abilities with the Google Assistant, Alexa or Apple's Siri?

Google, Alexa or Siri have not opened up all their application programming interface (APIs) and so integration at this point can be limited. Also, security is a big issue in banking. Would you say out your PIN aloud in public? So, I think the future of such integrations will be text and voice combined. 

Having said that, we launched a voice-based international remittance service to enable non-resident Indians (NRIs) to send money to any bank in India last year in November. With this new feature in the ICICI Bank’s Money2India app, an NRI customer can instantly initiate a remittance to his/her existing payees in India with just a simple voice command to Apple’s virtual voice assistant, Siri, on his/her Apple iPhone or iPad.

Also, we are working on vernacular languages. We are developing an application for Hindi users called meraiMobile where we will use voice, text and maybe even pictorials for customers. This will not only offer a simpler experience but will also help us acquire more customers.

How are you training your machine learning algorithms?

Banks have huge amounts of structured and unstructured data. What we are doing is using machine learning to draw inferences out of that data, in other words business intelligence. BI can be used for fraud and risk modelling and shaping customer service modelling, among other things. 

However, if you are also hinting at the use of AI in rationalising complete decisions such as loan approvals and wealth management, we are a little far from automating the complete process. There is still human intervention needed. Several foreign banks have tried doing the same thing with a software called Amelia but had no huge successes. 

How are you using a combination of these technologies to market your products?

We do have context-based offers and services. We can use a combination of technologies to push offers and services for digital as well as physical stores via our cards or loans. We recently announced an association with Paytm for e-commerce services. 

We are also offering deals in the physical stores. For example, if you have given permission to the iMobile app for your location, then you will get offer alerts if you happen to enter a particular store we are partnering with. We are also working with companies like CreditVidya, which offers tertiary information helping us get a customer's credit rating.

You said that ICICI bank has invested in a few companies. Could you name them?

As I mentioned before, to collaborate, co-create and nurture fintech companies or startups, the bank has included many innovations in its offerings. We offer two centres, one each in Mumbai and Hyderabad spanning 6,000 sq ft and capacity to seat 100 individuals overall. 

The bank also has a corpus of Rs 100 crore for taking a stake in promising fintech companies or startups. Of this, Rs 40 crore has already been allocated for investment. 

We have worked with Signzy for digital KYC, and helping individuals and institutions open current accounts in a matter of hours. 

Another product co-creation was done with FingPay, which enables merchants to accept digital payments through biometric authentication using Aadhaar Enabled Payment Systems. It is used in ICICI Bank’s Eazypay app. We have picked up a 10% stake in the company.

We also invested in Arteria, which offers supply chain solution on cloud and payment integration services to original equipment manufacturers and their supply chain network. We are working with them to grow the loan book for the channel partners. An agreement has been signed for taking a 10% stake, though the amount has not been disbursed.

You mentioned about the bank working on blockchain projects. Could you share some details?

ICICI Bank first used blockchain in October 2016, to execute transactions in international trade finance and remittance in partnership with Emirates NBD, a leading banking group in the Middle East. This April, we announced that we have brought on board more than 250 Indian companies on its blockchain platform for domestic and international trade finance. 

The usage of blockchain technology simplifies the paper-intensive trade finance process by bringing counterparties on the same platform, enabling decision-making in almost real time. It also eliminates the need to wait for courier of paper documents across cities or countries, and verification through trade intermediaries. 

At home, we have worked with an open-source blockchain tech called Stellar on which we have run pilot projects trying to execute cash transactions. We are also working with other banks on BankChain.

How have emerging technologies helped the bank?

The use of emerging technologies has been very positive for the bank. Our cost-to-income ratio has been coming down steadily after remaining flat just after we started expanding on our tech detail. On the retail part of our business, the cost-to-income ratio stood at 38.8% in the first half of 2018. 

More interestingly, the bank has grown its operations and customer base hugely in the past few years, yet we have maintained our headcount. This has been possible because of the use of technologies to complete simple tasks that would take up our staff's time. Now they can work on processes that are fairly more complex.

---