ICICI Bank leverages iPal chatbot for customer service
ICICI Bank’s AI powered iPal chatbot has interacted on the bank’s website and mobile application with close to 3.1 million customers, addressing 6 million queries with nearly 90% accuracy in eight months (Feb – Sept 2017)
ICICI Bank’s AI powered iPal chatbot has clocked 6 million responses till date with a run rate of 1 million chats per month. “Interestingly, the numbers are on par with the volumes hitting the call centre,” says Madhivanan Balakrishnan,Chief Technology Digital Officer, ICICI Bank. The chatbot, now on a learning curve, handles customer queries on the website and mobile banking application, and offers instant resolution. ICICI claims it is the only bank in the country to offer chatbot services on its mobile banking application, iMobile. In addition to offering instant responses to queries, the iPal engine on iMobile undertakes financial transactions as well – an industry first feature. It enables customers to undertake financial transactions like bill pay, fund transfer and recharges. iPal is also available on Pockets, a digital bank on mobile phone, which also incorporates an e-wallet.
“We have taken a few initiatives in the AI/ML space, which will be scaled up from two perspectives – customer experience and efficiency and productivity improvement of the employees. Thus these technologies will help both the customer and the bank,” says Madhivanan.
iPal was rolled out early this year on two of the primary applications- ICICI Bank website and iMobile application. The iMobile app is used by 6 million customers with a high ratio of active usage. Besides the funds transfer and accounts view, the most viewed features, the next biggest is the iPal chatbot used for services. The bank is also working on a Siri like Natural Language Processing (NLP) query resolution mode. The customer just has to speak out the query, which will be understood by the iPal engine.
iPal is almost like interacting with a live operator with an option to switch over to chatting with a live operator. The best part of the process is, the chat bot learns from the live operator and applies it when a similar query arises the next time. Going forward, additional services will be added in the chat and NLP mode. The bank is building a much better responsive website. Every search query by the customer is taken over by the chatbot, which collates it, learns and takes the customer at the destined location with fewer clicks. This is being done with a fair amount of accuracy, which keeps on enhancing.
“iPal has three primary modules. The biggest value lies in the core engine of the iPal chatbot,” says Madhivanan. As it evolves, the engine will give greater capability to add more customer services on the iPal engine. The first, is the FAQ module. It helps to build the existing repository of knowledge and structure such that the chatbot is able to pick the relevant query and pick responses based on the questions asked.
API is the second module of the iPal chatbot. Query resolution also requires the chatbot to interface with the right system at the right time because the effort is to make the system completely open. APIs are specifically driven at fund transfer, bill repay, recharges, etc. “As more and more services are build, I do not see any difference between separately designing an application with an API gateway and automating the entire process, as it is being done, currently,” says Madhivanan. It’s important to understand what the customer wants, zeroing in on the problem resolution and then the backend picks up the necessary APIs or triggers the necessary process flow and gets it rendered. A big chunk of API rendering, configuring is managed in house by the bank.
The feature discovery module is the most interesting. Wherein, the chatbot itself is helping to navigate through the huge reams of data with the bank and forcing to think about patterns, tracks, which then pushes the bank’s team – be it products, services or operations to think about solutions based on the tracks and patterns e.g. queries regarding GST and demonetisation are lined up on the upper deck even before the usual and frequently asked queries, after the events occurred. The live operators or the call centre handle these queries for the first few days. Thereafter, the iPal takes over based on the learnings.
ICICI Bank has roughly 200 services on iMobile and close to 350 in the Retail Internet banking (RIB). The services that are the most conducive to an AI based logic are segregated and given to the Fintech companies / internal teams to work on them. Post that, an RoI story is build around it. There has to be a minimum number that has to be established. The 80 / 20 rule works most of the times. Twenty percent of the features have eighty percent of the relevance for the consumer and thus for the bank.
The iPal engine sits on the server side. All the security provisions that applies for the iMobile app also applies for the engine. Soon, NLP will also be enabled for the chatbot.
If you have an interesting article / experience / case study to share, please get in touch with us at [email protected]