How HDFC Bank is banking upon AI across the organisation
Use of AI in customer services has become mainstream, especially in the banking sector. HDFC Bank has been exploring an array of domains where AI can be deployed
The application of Artificial Intelligence (AI) in the banking industry is multi-functional in nature. There are numerous use cases, encapsulating the deployment of AI as a technology. For instance, customer service is a big area where AI has been deployed. Similar is the story of HDFC Bank, which had launched its AI based chatbot – OnChat – on Facebook Messenger in 2016. Within one year of its launch, the chatbot, created in partnership with Niki.ai, had marked 160 per cent month-on-month growth in transactions. Until April this year, more than 300,000 consumers have interacted with HFDC Bank OnChat and the value of transactions is close to Rs 2.5 crore.
Another example of use of AI in conversational banking is HDFC Bank’s virtual chatbot ‘Eva’ on Amazon Alexa and Google Assistant. Eva has answered more than five million queries from around a million customers with more than 85 per cent accuracy. Eva holds more than 20,000 conversations everyday with customers from across the world.
“We are attempting a lot of use cases across various functions. While OnChat is a well known example, we are also piloting various other use cases. In the HR space, we have used AI for initial screening of candidates in the hiring process. We have a roadmap which captures what we want to achieve across multiple functions within the bank. In order to establish the capabilities, we are going through Proof Concepts (PoCs),” says Nitin Chugh, Country Head – Digital Banking, HDFC Bank.
Multiple use cases
In addition, HDFC Bank has also made several deployments in risk management, credit scoring, employee engagement and onboarding, portfolio management. Besides, there are also good applications of AI from marketing standpoint. For example, AI can help the bank optimise and interpret campaigns in an optimised manner.
Commenting on whether there’s a need to present a business case to the top management for investing more in AI, Chugh informs, “At HDFC Bank, we use a combination of top-down approach, which is led more by strategic intent. Our disussions are not confined to RoI; however, it is an important consideration among other factors.”
In the area of security, HDFC Bank has recently completed a pilot for AI based Cyber Security Operations Centre (CSOC). The log data from CSOC is put for processing on the AI solution having big data capabilities and it was done for about eight months on a cloud platform. The AI solution will help in monitoring insider threats.
The bank charted out and has embarked on an articulated path for its AI journey. In the initial stage, the bank’s focus is on helping its workforce with AI assistants. This is largely information-led, but it could be related to workflow as well. For example, in the case of hiring, AI helps the first-level interviewer in screening of candidates. In the second stage, HDFC Bank will get to a point where some of the workflow will get signed to AI applications. Whereas, in the last stage, some workflow will operate autonomously – this will also use machine learning in a big way.
Chugh shares, “We are going to move in a gradual manner, because the entire organisation has to be prepared. In terms of skills, we are already exposing our workforce to these things. We are encouraging people to learn and are also organising a series of sessions across the bank. The familiarisation efforts are more directed towards people who will need to specifically use AI in the future. For example, the back-office team, credit team, products and portfolio team. We are unlikely to hire more data scientists in the future, because we may not be able to convert everything into actions. Undoubtedly, we will be producing insights and bringing out experiential knowledge, which will be supplemented by algorithms. On the analytics side, we are already a heavy user of algorithms.”
The key factor is about how the technology is applied, instead of how many people are needed. For instance, one may have a lot of data scientists for churning out insights; however, one should also possess the skillsets to turn insights into actions. “We may not need too many data centres, because as machine learning settles down, there will not be a need for too many people for data mining,” he adds.
The bank feels that AI will not replace any function entirely. Citing example of customer service, Chugh reaffirms it won’t be completely transformed and replaced by AI, because the function is customer-dependent. Likewise in terms of sales, several tasks can be done with AI; however, AI won’t take over the distribution function completely.
“None of the functions will get completely replaced by AI, but will definitely get transformed. In the future, they will also get transformed with real-time accuracy. To accomplish this, people will have to be assisted by AI,” he says.
Already having a Centre of Digital Excellence (CoDE), at present HDFC Bank is keen on establishing the domain of digital excellence, which involves a combination of five-six different practices such as social media related learning, research based knowledge, ability to track real time insights, etc. The bank has 25-40 use cases in the PoC stage, which marks a growth of 50 per cent from last year.
“We don’t need to establish too many use cases; we have a clear strategic intent, and we have identified the areas where we want to apply AI. We just need to test a few things in every domain where we intend to use AI.
Furthermore, the bank is also passing its learnings to other companies within the HDFC Group. A good example of this is OnChat. HDFC Bank started it with commerce transactions, and has recently added the stock trading
option as well, thereby complementing HDFC Securities. It has built another interface for the general insurance and life insurance products.
If you have an interesting article / experience / case study to share, please get in touch with us at [email protected]