Nascent AI heading towards Cognification
Banks and financial institutions speak of AI as the next frontier but at present its uses are just peripheral though the scope is infinite
While AI is increasingly becoming the next-in-line fad for the banking community, right now its uses are confined to chatbots, IPAs and similar interactive tools. What experts feel is that unless NLP, or Natural Language Processing, is perfected and deployed, there will be only peripheral use of AI in the way banks interact with its customers.
It is perceived that AI will have applications and better uses in areas like personal financial management, advisory services, advertising, recommendations to customers and security systems. Kasisto, the US firm that develops conversation platforms for banks to add virtual personal assistants to their mobile and tablet offerings, has said that it is developing an omnichannel product that will effectively function as a ‘banking brain’ that can have contextual conversations with customers based on what it knows about the customers, making use of customer data to create more personalized conversations. The thinking is shifting to technology that can be applied to a variety of channels. Mastercard and Standard Chartered Bank are the initial users of this product.
There is this new process of Robotic Process Automation, or RPA, which global banks are increasingly using recently, mainly to reduce costs and move from services through labor to services through software. It involves the creation of virtual workforce, which has helped these banks to sometimes even eliminate human intervention in decisionmaking, in execution. Some banks in the west, which have been using the services of lawyers to finalize loan agreements, are now using AI systems, which have not just cut down costs, but have also reduced the level of errors resulting from human judgment.
The prediction is that AI will be able to independently analyze what is there in the whole of digital world, meaning the internet, apply this to internal data and create intelligence based on which solutions are created.
“Adoption of AI is no longer an option or a showcase item. Major banks are already
experimenting with AI solutions in some way or the other and are now solving real problems with it,” says Anup Purohit, CIO, Business & Digital Technology Solutions at YES Bank. “Early versions of chatbots serviced customer FAQs. However, more and more banks are now adding mainstream banking services into chatbots and are integrating their AI solutions into various chat applications. There is a definite push by banks to increase the usage of their AI solutions among customers. Banks are also using AI as enabler for their employees. They are empowering their employees with chatbots that answer process and product specific queries thereby bringing down processing time and increasing customer delight. AI has already made inroads into the BI and cyber security domains. There are live examples of ML solutions analyzing network traffic to identify potential rogue connections,” he says.
Sanjay Sharma, head - Technology, Innovation & Customer Fulfilment at RBL Bank, believes that each bank has had a unique journey in AI, depending upon its needs and growth strategy. However, the key application has been customer facing chatbots, answering FAQs from customers, conducting basic customer transactions, allowing banks to rationalize costs and to help its staff devote more time for value added work. “A few banks have been experimenting with other AI driven customer facing applications such as assessing customer feedback in real time, offering deep customizations based on customer insights generated across social media. There have been a few instances where banks have tried to conduct pilots in intelligent back end automation as well, as a next step of Robotic Process Automation,” says he.
He also maintains that banks are currently testing many AI driven applications, with adoption of a few for mainstream consumption. “The year 2018 could be a year where AI experiments in fintech/ innovation labs of banks would be consumed in the banks’ businesses on a mass scale. To continuously enhance AI capabilities and help in achieving the growth, it is necessary for Indian banks to develop internal AI capabilities and formulate a welldefined AI strategy,” he adds.
However, Kalpana B, partner and headRobotics and Cognitive Automation, KPMG in India, does not think banks and financial institutions in India are making use of AI in a worthwhile manner. Says she: “Well frankly, I am not aware of any AI applications in use by Indian banks. I am not disregarding the conversational chat bots, but these are still at a very nascent stage. The conversational chatbots are yet to be advanced to a degree where it can support beyond predefined queries.”
She says currently the adoption of AI in banking is at its nascent stage. “To be honest, we are in need of useful data in order to train anything in intelligence. I see a rapid catch up in the next three years to avoid existential threats,” she adds.
According to Rajiv Kumar, MD & CEO, Universal Sompo General Insurance, a host of AI-related technologies such as NLP, Computer Vision, robotics, machine learning and speech recognition have surfaced in the industry and have substantially progressed over the years to coalesce into Indian systems that are capable to perform activities such as think, learn and continuously adapt. “However, at present, the easily-configurable and implementable solutions using AI, such as chatbots, personal finance trackers and advisors with the rule-based configuration to resolve policyholders’ queries are in use and are also customer faced in the insurance segment,” says he.
AI is gaining on its capabilities at a fast pace facilitated by better computing power and high caliber hardware. Alongside, robots are gaining foothold, replacing human labor in critical areas. Chatbots and IPAs are not the real domain power of AI. Purohit says AI is defined as the ability of a computer/software program to perform tasks normally requiring human intelligence. “Human intelligence relies on past data/memory, current data points and at times impact of a decision on the foreseeable future to provide the relevant answer, the best-alternative or a decision. AI solutions are increasingly becoming better at mimicking this human behavior at far greater speeds as compared to humans. Besides, they do not have memory limitations like humans. This would be the mainstream use for AI – augment human decisions with ‘best-alternative’ data points and going a step ahead, decide on behalf of humans,” says he.
He adds that while AI as a platform can cater to any kind of industry, fintechs are taking the effort to develop the libraries to make their AI solutions banking-aware. They are also creating libraries specific to certain markets/region/country. This allows for faster adoption of the AI solution and avoids discovery of mistakes over longer periods of time.
“Banks too are getting innovative and implementing AI solutions in a novel way. For example, use of vision based AI solutions can analyze video feeds from ATMs in real-time. A person standing too close to the ATM, bending down too much or making unusual movements can be tagged as suspicious. They can trigger an alarm or disable the ATM till the service provider scans it for snooping devices or pilferage,” says Purohit.
Sharma is of the view that banks
need to expand boundaries of AI beyond conversational analytics. He says a few fintechs are offering applications focused on CXO dashboard for internal consumption, which can be leveraged. A chatbot can answer a few critical questions such as which branch is likely to achieve maximum growth in terms of new business in the next quarter and which branch is going to lag; what are the skills, which are in the highest demand in the market; how have the last 5 general elections impacted banking; are there any measures successfully adopted by banks in other regions: and what is the % adoption of mobile payments in tier II cities and beyond.
“There are a few innovative uses of AI in fields such as fraud detection, AML, compliance, automated audits, etc. which hold good promise,” he adds.
For Dr N. Rajendran, CTO, National Payments Corporation of India, while AI will find applications in business analytics and fraud detection systems, it will be real time analytics with AI and ML tools that would improve the business decisions, fraud prevention, cyber security and of prediction in these areas.
Rajiv Kumar says applications such as wearable devices, telematics and chatbots capable of query resolving hold promise for the insurance sector. “As you know, IRDAI has recognized the role of wellness aspects in risk assessment and product design for health insurance offerings and use of telematics for the motor segment. The insurance segment is gradually evolving on these fronts,” says he,
The prediction is that the next stage of digitization and automation will be the process of ‘cognification’. It is described as the combination of IoT and AI. It is believed that the synergistic effect of applying AI to connected devices will be incalculable. Purohit believes this will be the natural progression. “Currently AI systems exposed to users are managing low risk queries or processes. As the confidence in these system increases with more usage and more learning data, users will compare their decision with the one provided by the AI system. At some point, we will allow the AI system to make these decisions,” says he.
The best example would be the traffic navigation applications. A large percentage of the users completely go by the ‘shortest travel time’ path provided by the application, he argues.
Sharma says a few specific actionable items to achieve cognification are:
Select the next best action
Achieve N=1 level customization
“AI will help analyze data across multiple unstructured, semi structured and structured sources and generate meaningful and actionable insights to offer customized and differentiated solutions. This should help a bank to become more cognitive - a bank that thinks and insights driven over the next few years,” says he.
Kalpana too believes that AI will be used with processes to make specific decisions. “I think it is very much possible. There is a continuous evolution that is taking place in the technological domain and AI has definitely made its way from sci-fi movies into our daily life decision-making,” she says.
Rajiv Kumar explains that cognification involves advanced technology platforms that can address complex situations that are characterized by ambiguity and uncertainty. He has words of caution: “Cognitive computing has begun to augment business decisions and power performance right alongside human thought process and traditional analytics. However, there are a good number of industry leaders who would want to see cognification in the area of performance management but for me, the role of human decision should always be valued more than machines and there should be some guidelines or upper limits to use AI.”
FINTECHS AND AI
Purohit points out that banks today are increasingly partnering with fintechs through their co-innovation labs. “Fintechs, by their very nature, focus on solving a specific problem and/or creating a specific product within short timespans,” says he. “They mostly tend to use the latest technologies and are therefore adopting AI in a big way in their solutions. Which means banks now have easy access to the latest AI solutions along with the necessary skillsets and knowledgebase to implement it.”
He goes on to add: “It’s a win-win situation for both. Banks get to implement and use the latest AI solutions in short timespans while fintechs get to finetune their product by analyzing the huge datasets available with the bank and feedback from domain experts in the bank.”
Kalpana explains that fintechs are contained microcosms to start change in the sector. They normally cater to a specific need or use and because of their extreme focus, they help banks, says she, adding: “For a variety of tasks, it’s like orchestrating fintechs across a process /processes and no one has yet completed the journey. There is a huge difference if you are an individual violin/cello player versus if you are one amongst many in a philharmonic orchestra. Sometimes we observe the size of organizations inhibits speed and progress.”
Sharma says fintechs and challenger banks have taken a lead in developing and consuming AI driven applications and banks have been experimenting with such applications through hackathons and accelerator programs. “I see banks and fintechs collaborating over APIs, sandbox for developing innovative AI driven solutions,” he says.
Dr Rajendran too says fintechs and
challenger banks have taken a lead in developing and consuming AI driven applications. He also mentions that in the days to come adoption of AI will increase in the banking and financial services sector.
Is there a prospect of the use of AI leading to replacement of humans in certain operations? Or does it serve the role of ‘augmenting’ the operations?
Says Kalpana: “Yes, there are possibilities where AI may replace humans if that serves the objectives better. And yes, it would augment the human capability in some other context. For example, a contact center interface could be eliminated if AI can predict what may happen with a customer and pre-solve their potential cause to call the center. At the same time, it can also augment a personal relationship manager to better serve his clientele by understanding their life stage /create and review complex algos and suggest what investment best suits.”
Purohit too says AI will augment and will definitely help free up human bandwidth in certain operations. And this freed-up bandwidth will be redeployed towards new product lines/services and reach out to a wider customer base, he adds.
Stating that there has been a lot of talk and debate around this topic, Sharma argues that AI should be viewed as a mechanism to help a bank become more intelligent and smart. “This is possible when employees can focus more on value added tasks, while AI based applications help to automate repetitive, time consuming tasks. I think at least for the next few years as banks will augment their AI capabilities, AI will act as the employee’s companion to generate results better and faster and deliver better.”
Dr Rajendran says AI will be used for automation in certain operations and it may reduce the humans in certain operations.
In this context, will it be a challenge for banks, increasingly using AI, to retain the human touch?
“This question has come up in the past,” says Purohit, “when banks launched ATMs, internet banking, IVR banking, SMS banking, mobile banking. While they reduced or eliminated the customers’ branch visits, these options provided convenience and time flexibility to the customers. The AI option is no different. The benefit it will bring in will far outweigh the loss of human touch, if any.
“I think humans would do what humans do best – connect with others,” opines Kalpana. “I am not seeing last mile adoptions even in the western world yet. We need to be mindful of future possibilities and prepare rather than fear on how to retain control. Because, when we fear we freeze and that may be dangerous for the organization.”
Sharma asserts that the role of AI is not to replace humans but to augment their capabilities. Most banks are expected to adopt a hybrid approach, he says adding “AI driven robots and personal relationship managers would co-exist to support the customers.”
Rajiv Kumar says since insurance is driven by data, use of AI will have a huge effect on the company’s bottom line and the satisfaction of the customer. “However, in advanced nations, AI has revolutionized the way insurers gain information from their customers but from an Indian perspective, the consumer is not that much aware of new technologies. Hence, for an insurance company, while there is growing demand from millennials for technological advancement, our rural, semi-urban customers, senior citizens still require the human touch. So, this will be going to be challenging for us, as, while on the one hand we have to increase our investment for technological developments while on the other, we still have to retain our traditional methods.”
He says few large insurance companies in India have initiated to utilize AI for either sales or customer support. “In my view, the operating and business models of insurers are evolving. Technology trends such as AI, machine learning, blockchains and robotic process automation have significant potential to streamline insurance operations and enhance customer experience,” says he.
Sharma says going forward, AI can offer conversational analytics as one of the many streams. “With increasing instances of cyber frauds, AI could be significantly deployed for assessing suspicious transactions, pattern recognition and protect the confidential customer data at an early stage. Other mainstream uses of AI could be seamless transaction flow, better KYC, KYB and assessing the customers with highest probability of sales conversion.”
Kalpana sees in AI the potential of virtually changing the banking model and disrupting the way traditional risk and reward matrices work in banks. “In my view, if we imagine the current back and look for AI cases, we may not do justice to the art of the possible. The key therefore is to reimagine the future,” says she.
Dr Rajendran says AI is used for various requirements of data analytics and for acting as digital interface for interaction with various systems and end users.
Rajiv Kumar foresees the evolution of enterprise AI solutions, which can enhance operational efficiency, improve time-to-market capabilities, enable a more intelligent way to sell and service customer and more. “However, from an insurance perspective, the role of AI would be in customer servicing, underwriting and settling claims,” says he.
Specifically, what about the possible application of AI in areas like AML, algorithmic trading, fraud detection, customer engagement, risk management?
Says Kalpana: “Well, AI can be used across most of the areas that you’ve listed out. I can give a few examples to show how far they have been applied in a few of the areas. In the area of Algo trading, AI is used for developing trading strategies and suggesting portfolios to clients. AI in fraud detection is seen in the form of reduced time to scrutinize every financial transaction. Customer engagement is enhanced thoroughly with the application of AI, by making it superbly interactive and personalized. I suspect that AML screening would be quite different if AI is applied and it has to work in consortium across banks and regulators and not just one.
Sharma is of the view that priority for banks will be to provide a more secure transacting environment for its users. “For the same, we can expect AI to be deployed for: assessing the real owner of the transaction by analysis of the transaction sources; Know your Business (KYB) to detect sources of income, nature of transactions for a strong AML and fraud risk management platform; velocity checks, taking action on its own when anomalies are detected; automated audits to dig out generally unforeseen patterns; actively hunt for domestic and global regulations to ensure 100% compliance; deliver insights, market news on fingertips; significantly drive customer loyalty through sentiment analysis - assessing the mood, tone, emotions of the customer through natural language processing capabilities for offering deeply customized solutions for that particular context.”
AI & CBS
Has AI got the scope to be an integral part of CBS in banks?
Says Purohit: “Rapid developments in AI space will make it very easy for CBS system providers to bundle AI into their products. Many will invest in developing their own AI solutions. However, I foresee CBS players providing capabilities to integrate with other AI solutions rather than bundling one. Banks have already started their AI journey and with every passing day, the AI solution is getting better and speaking with more and more systems in the bank. Switching over to a bundled solution will be time consuming and may not be worth the effort.”
Kalpana upholds the dictum that for any investment, there should be a why. “I would think of it in a different way. Why would you get in the way of a legacy Core Banking Solution, when you can make an independent AI-run digital platform altogether, which can take things smoothly? Eventually, the CBS can be replaced with an AI digital setup. Imagine the pre-CBS era and why we got there. Again, reimagine than renovate the existing.”
Sharma thinks over a short term, banks may focus on customer facing and security related applications for AI deployments. However, he is certain CBS is one area which is ripe for innovation. Already, banks are experimenting with cloud based core of the future with open standards of coding, simpler stack, etc to enhance performance.
Would banks appoint chief AI officers?
Purohit says while such a title may not have been given to anyone now, banks already have a person or a team performing this role. “I believe banks have understood the need to have an enterprise wide AI solution and are in their different stages of implementation. It’s only a matter of time before a title is announced officially,” says he.
Sharma feels the nomenclatures and titles may vary with each organization, but the role would definitely require a strong understanding of the uses of AI for all internal and external customer facing applications. “One important aspect to watch out for is the level of such an officer in the organization - whether he/she is a member of technology transformation team or rolls up directly to business head or to the CEO & MD of the firm. This will indicate the strategy which the bank has adopted with regards to consumption of AI,” he says.
Kalpana has the last word: “I think we are again making the mistake of imagining what the roles on existing structure are. We would have CEOs to the last employee in hierarchy utilizing this technology as appropriate. We don’t have a chief MS excel officer, in jest, so why only for AI. We need a collaboration of diverse skills and my view is by putting one person in charge, we would run risks.”
Anup Purohit believes mainstream use of AI would be to augment human decisions with best-alternative data points
Sanjay Sharma points out that fintechs and challenger banks have taken a lead in developing and consuming AI driven applications
Kalpana B believes use of AI is in a nascent stage and there is need of useful data in order to train anything in intelligence
Rajiv Kumar foresees the evolution of enterprise AI solutions, which can enhance operational efficiency, improve time-to-market capabilities
Dr N Rajendran feels it will be real time analytics with AI and ML tools that would improve business decisions, fraud prevention, cyber security