AI is one of the most talked-about technological innovations of our times. Everyone in the industry is talking about AI, automation and deep learning. However, very few know what AI is and how it works. Possibilities for AI are limited only by one’s imagination. There are a few underlying principles which if followed provide a solid foundation for a futuristic solution. Let’s dig into this a bit more.
A lot has been written on what banks can do with AI, the various use cases, and the wishful end states in various functions. Many banks are doing something or the other about AI but very few seem to have a solid progressive strategy to fully utilize it. Most of the banks are doing it out of compulsion. They have automated many manual processes and have introduced various stand-alone pilot programs but have very less to show in terms of the gains achieved.
There are banks which have implemented chatbots remarkably well to enhance customer servicing. Also few banks seem to be getting targeted offers pretty spot on. But to have a sustainable competitive advantage through the use of AI, banks need to do a lot more and a lot faster.
There are a few crucial points which if done correctly can lay a strong foundation for the bank of the future.
Raw and unprocessed fuel cannot be used to power even a small bike engine. Similarly, raw and unstructured data cannot be used effectively to power the AI engine. It is true that banks have a treasure of customer data but it is worth examining whether they are readily able to use the same to power AI.
One of the major issues with banks is that data is stored in numerous systems and more often than not the systems do not talk to each other and have their own version of data for the same attribute. My own bank has stored different mobile numbers for my savings account and credit card account. As a result, I keep getting cross-sell messages on both the numbers. This is just a small example but as the number of systems increases, the complexity and hence the magnitude of problems caused by data inconsistency increases. Though this issue is not fatal, it limits the application of AI to a great extent.
Another issue with data is integrity of data even in a stand-alone system. Example – A bank created a data model to cater to the prevailing regulations. Sometime later, the regulations were changed and it called for an introduction of an extra field. As the time was limited, the operations team decided to capture the new piece of data in the remarks field. A few years later, auditors had a hard time figuring this out. AI is still not that intelligent to decipher such human hacks.
So, to reap the benefits of AI in the long term, banks need to start getting their data in order at the earliest. It will need some investment and unfortunately there is no shortcut to fix this.
AI will be self-sufficient one day. But till that day arrives, it will need help from humans. There are two levels at which humans need to intervene in AI systems to ensure business continuity and to improve the system.
Basic level of human intervention is needed when the AI system runs out of options and looks up to us. For example, a financial advisor chabot may encounter a unique situation which needs human intervention (at least for the first time). In such cases, the chatbot can handover the conversation to a bank executive to solve the problem or can raise a service request to be catered by someone at a later point of time. Banks need to be ready with such fallback mechanisms while implementing AI solutions.
The next level of human intervention is needed to make the systems more intelligent. The AI systems (machine learning systems to be specific) observe human behavior and convert that into a mathematical model to emulate it the next time. While these systems are able to learn, to make they are sustainable, they must continuously be taught new models by a human being. That human being is the data scientist. Banks need to plan for such maintenance and enhancement activities while creating a solution.
AI can be very aptly compared with digital transformation, in the sense that everybody wants to do it but very few know how to do it.
Such a thing cannot be done as a separate function. It cannot be a plug and play silo which will work. Each and every system, every business process and all channels need to work as a unified machinery to deliver truly digital service or AI powered service for that matter.
Business leaders, technology leaders, operations managers, everyone needs to be sensitized about the importance of AI enabled services. These leaders need to communicate with their teams and assure them of how AI can be beneficial to organization and to an individual. Fears of employees must be allayed. I can say from my personal experience that driving even a small cultural change in a bank is a herculean task. Unless everyone embraces this with an open mind, the implementation is going to be very difficult.
One additional factor is to find talent with the right mix of expertise in data science and banking domain. There are very few resources available in market today having both these skills. Banks can also look at training their high aptitude employees in AI methodologies.
We talked about data consistency across multiple systems. We also talked about the importance of thinking about AI as a way of life across modules and not as an afterthought. But who creates these systems? The answer is – software service providers. Banks are dependent on these partners for most of their system needs. Very few banks do end-to-end in-house development. These partners are in a good position to help banks implement AI related services. An ideal partner should have a futuristic roadmap but at the same time be accommodating of the banks’ needs, limitations and should come up with appropriate solutions. Similarly, banks must focus on long term investments to move in the right direction. What matters is being on the right track, no matter how slow the start.
All the points discussed above do not mean that banks stop thinking about various use cases and just focus on these foundational activities. Both should go in parallel. At an application level, banks should engage in pilots with whatever data they have, but at the same time should not miss the bigger picture. At the strategic level, banks should plan to improve the quality of intelligent services in a phased manner as and when the back-end support (technical as well as business) keeps evolving. Though we have been talking about this for a few years now, there is still a very long way to go and I indeed believe that we will collectively achieve the wishful end states that we have thought of.