Not long ago, banking was synonymous with long queues, pass book entries, cashier tokens, bored security guards and woefully ineffective air conditioning. Many would argue that it’s still the case, especially for the last two points, but there is no doubt that banks are rapidly modernizing behind-the-scenes.
An example of this would be ICICI Bank’s “fully digital banking” adverts -- which I endured during the recent India vs England cricket series on Disney+ Hotstar. Everything from opening a bank account, applying for a credit card, submitting KYC documents, and meeting bank executives -- all done online, through the smartphone app no less. Sounds too good to be true?
Deekshith Marla and Afaq Hussain don’t think so. In fact, in many ways, they are the joint architects of the silent technological upheaval happening inside some of India’s largest banks.
AI makes everything better
Deekshith Marla is the Co-Founder & CTO of Arya.ai, Mumbai-based startup that builds AI applications with deep learning algorithms to automate and speed up a lot of tasks in any process, while Afaq Hussain is the President of Global Strategic Alliances and Enterprise Business at Intense Technologies Ltd, which designs and builds enterprise software suites and applications for some of the largest companies in India and the world. Both of them are ISVs (independent software vendors) that are intimately involved in India’s banking and finance sector, not at all surprised by the incoming digital banking wave.
In fact, according to PwC India’s December 2020 report, 70% of Indian enterprises have implemented AI in some way or the other in their business process, of which banking and financial services is one of the highest AI-enabled sectors with 82% process enhancement.
“Banks are leveraging AI through chatbots and voice assistants at the front end to mimic live employees, and everything from customer identification and authentication to delivering real-time communications,” according to Afaq Hussain, who says customer service is the highest contributing segment for the proliferation of chatbots and virtual assistants, with a 43% banking industry market share.
Echoing Afaq’s assessment, Deekshith Marla says, “AI chatbots span across a wide category and isn’t restricted to just conversational bots. It also includes many process automation where various services are provided via integration with applications like WhatsApp.”
With an increase in the scope of tasks that can be automated in a banking environment, Deekshith estimates, “AI chatbots handle almost 70-80% of first-level customer queries and enquiries -- everything from creating an account to uploading KYC and getting statements and current balance, a person doesn’t need to visit a branch or even an ATM for these things anymore. Visiting a branch or calling a customer care number is now necessary only for rare scenarios or for expert assistance.”
From smart processing of cheques
In most Indian banks, Deekshith mentions how certain tasks related to cheque processing are automated -- reading the MICR info or account number on the cheque, for instance -- and that’s where Arya.ai’s cheque automation product fits in. “We automate all inward and outward cheque processing levels while reducing human intervention, and design workload to only focus on the percentage of cheques where AI may fail so it can be flagged to appropriate bank employees. At a couple of banks where our product is deployed, daily at least 40% of cheques are completely processed without any human intervention,” he says.
The application that Deekshith’s Arya.ai startup has built scans a cheque deposited for processing at any bank and cross-checks everything from name, date, account number, amount in numbers, amount in words, signature, and other bank-related information that depositors write down on the cheque manually -- so a significance chunk of the solution is OCR (optical character recognition) and image processing which includes matching the signatures on cheque and verifying it with the signatures saved against the account in the bank’s internal database.
“Where it would take a bank employee anywhere between 30 seconds to 1 minute to manually verify every single cheque in the past, our application processes 18 cheques per second,” according to Deekshith Marla. That’s no small feat.
“Our signature validation accuracy is at 94% right now,” Deekshith tells me, which is pretty much the highest number anyone could attain till now. For the AI model to achieve this feat it needs at least 15 lakh cheques to train on, data which is provided by the banks.
“A large private sector bank in India used to get four lakh cheques in a day, which they would take about eight hours to process manually and clear overnight by submitting them to RBI or NPCI’s systems, and we have brought that time down to just two hours,” says Deekshith Marla, further elaborating how if the system isn’t confident of processing any single cheque automatically -- because of illegible handwriting issues, for example -- it still has the option to flag it to a human banker for manual processing. The cheque processing application that Deekshith helped build at Arya.ai is now deployed at one of “India’s largest public sector banks, helping clear eight lakh cheques per day” in his own words.
To speeding up KYC verifications
With respect to digital KYC, “AI and ML play a key role in areas of face recognition, fingerprint recognition, digital onboarding of a customer, among other things,” mentions Afaq Hussain of Intense Technologies Ltd, further explaining how a “greenfield Indian telecom operator” was able to take advantage of this to scale from zero to 100 million customers in under six months -- no prizes for guessing which Indian telecom operator is being hinted at here (of course, it’s Jio).
The digital KYC based customer onboarding journey Intense Technologies helped create can be used in a variety of scenarios, according to Afaq Hussain -- from telecom to government citizen services, for banking and insurance applications, among others.
“We were able to bring down a customer’s onboarding time from six hours to 15 minutes and now less than one minute,” Afaq mentions, while referring to a banking scenario. In another example, Afaq points out, “Tax notifications which used to take up to 63 days for the IRS to send to its customers now take less than a day.” All of this is possible because Intense Technologies have AI-enabled applications that send out and manage over 200 million notifications in a single day, according to Afaq Hussain.
Constant learning and adapting to shifts in customer behaviour is where AI and ML drives efficiency, Afaq believes. “If a bank primarily communicates with a customer over email and the customer actually responds faster over SMS alerts, then probably reducing emails and sending SMSes first is an inference drawn by the bank based on artificial intelligence and machine learning capabilities,” he says.
“For loan requests, there has been an increasing adoption of AI usage, especially for KYC verification and the borrower’s general profiling. Automation of KYC verification offers a big improvement in customer experience and reduces the need for manual paperwork,” says Deekshith Marla of Arya.ai.
Both Deekshith Marla and Afaq Hussain agree that all of this is made possible by computer vision applications and AI-enabled pattern recognition tools that can be trained and continually improved through deep learning models. And their partnership with IBM, in terms of running their applications and AI workloads on IBM PowerPC -- particularly Power 9 and Power 10 systems.
Every developer needs a tech-savvy friends
Deekshith Marla, co-founder of a young startup like Arya.ai, probably is acutely aware of this fact more than Afaq Hussain’s time-tested Intense Technologies Ltd which is over 30 years old, even though both enable banks, insurance companies and other financial institutions to leverage IBM’s hardware and software infrastructure.
“You need high-performance computing infrastructure to run AI (artificial intelligence) and ML (machine learning) workloads, where each GPU -- like a V100 -- can cost $10,000 or more,” Deekshith highlights, which is challenging for both startups and their clients to justify at a proof-of-concept stage. This is where companies like IBM helps, Deekshith says, “IBM already has a good connect with all the leading banks in India, and when they saw our AI application they readily loaned us their Power 9 systems, initiated a dialogue for us with their banking customers so we could run the application tests inside the bank’s data centre for them to see process improvements and efficiency gains first-hand.”
Afaq Hussain also attests to this tried-and-tested partnership model that IBM (and other technology giants) operates to accelerate innovation across different business sectors. “The real work for us actually starts after aligning yourself to a customer’s vision, where you move beyond the proof-of-concept stage, and that’s where IBM has truly shined for us,” Afaq mentions, pointing out how the experience of deploying IBM machines in customer premises, startup premises, to assisting Intense Technologies with testing and ensuring things are moving forward has been nothing short of brilliant.
In particular, Afaq points to the IRS solution deployment which impacts $12 billion of annual revenue and the other being the AI-enabled journey of one of India’s largest private sector banks, “In both these instances, the way IBM Systems team collaborated with us was just awesome,” he reiterates.
Ravi Jain, Director, Server Sales, India South Asia, IBM, tells me how discussions around AI and ML in company boardrooms have shifted from clients asking about proof-of-concept workloads and lack of meaningful startup case studies (back in 2017) to businesses rushing to drive process efficiencies and accelerating infrastructure digitization plans all through the 2020 pandemic hit year -- the need for which is only going to increase in the coming weeks and months, believes Ravi Jain.
Afaq Hussain echoes Ravi Jain’s views, while summing up what the future holds for digitization not just in the banking and financial sector but beyond, when he says, “As much as Covid-19 pandemic has disrupted our lives negatively, it has forced every company and business to digitize fast and in a smart way without burning a hole through their wallet, including your neighbourhood bank's branch to provide meaningful customer relationship journeys.” Which is just fine by me, to be honest -- no more woefully ineffective air conditioning to look forward to, thank you very much.