The age of AI in banking: reimagining financial decision making

The Indian Retail Banking space has gone through multiple rapid developments in the last 30 years. Unlike its European or American counterparts, banking in India isn’t broken: retail customers’ interests are protected. It is very convenient for a customer in India to open a new bank account. In fact, a majority of Indians already have bank accounts. With the advent of UPI payments, India has a far superior payment infrastructure than anywhere else in the world. Customers can transfer money in a jiffy, and get instant confirmation. Neobanks, therefore, face a large entry barrier.

So which problems can neobanks solve?

Speaking to any mobile phone user in 2006, one would’ve concluded that the user was quite happy with their mobile phone. But that didn’t stop Apple from completely reimagining the phone. They changed the way we use phones forever. Sure, we still use our phones to make calls; but the primary use case is now a much larger set than that.

Similarly, neobanks must reimagine retail banking to change the way we bank today. Banks today provide these primary services:

1. A value store: The ability to store money safely (investments also fall into this category).
2. Money movement: The ability to move your money safely.
3. Access to credit: The ability to loan money when you need it.

Neobanks still have to provide these services in a better, more customer-friendly way. It doesn’t stop there, though. They must also do something beyond these primary functions, just like Apple did with the iPhone. Question what is, and build something new. Something bigger.

What’s this something that lies beyond banking as we know it today?

Reimagine Decision Making

All financial decisions are quantitative in nature. These quantitative decision making problems are interlocked with customer’s emotions. Take the following questions for example:

1. How much should I spend in a month?
2. What can I spend it on?
3. How much should I save for the future?
4. How can I grow my wealth?
5. Is this affordable?
6. How much insurance coverage should I have?
7. How can I save more? Or can I spend more?
8. How can I manage my assets and liabilities?
9. How much would I need in case of an emergency?

We ask ourselves such questions multiple times a day. Most customers make decisions based on intuition. We are, after all, influenced by emotions. Very few try to make decisions rationally. And so, most of the time, these decisions are made sub optimally, depending on our financial state. This is where neobanks could step in. Providing actionable answers to these questions, tailor-made to each customer depending on their financial situation, would be a truly personalised experience. There’s scope to build personalisation in the retail banking space.

There are some fundamental problems one could face if providing such a service. Here are a few:

1. Most customers aren’t even aware of the questions they should be asking. They just don’t think like that. This is, then, also a behavioural change. Any product that tries to solve this needs to first make customers think about the questions they aren’t yet asking.
2. Even customers who know which questions to ask aren’t used to asking them. Or looking for services to find answers. And so, this is also about earning trust in addition to changing behaviour. A product that tries to fit into this gap needs to also earn trust. How? By answering clearly, in a way that’s specifically applicable to that customer.
3. There’s a small segment of customers who are aware of the right questions to ask, and who also seek professional services to manage their wealth. But the solutions that wealth managers offer are very limited in comparison to the kind of solutions that could be offered instead. Using a quantitative approach, technology and AI can truly personalise this experience. Add behavioural science to the mix, and you’ve got a product capable of bringing about the necessary behavioural changes in customers.

These fundamental problems need to be addressed to solve the larger problem of effective financial decision making. And this requires reimagining multiple product constructs.

Applied AI and Behavioral Economics lie at the centre of all of this. At Jupiter, we’re solving these problems fundamentally using the latest advancements in Applied AI and Behavioral Economics.

Take the example of PFM (Personal Finance Management). Here’s a 50,000 ft view of how Applied AI and Behavioral Economics can help solve it.

A classic scenario that many customers face is: should I pre-close my debt obligations (Home loan or Auto Loan etc) or should I invest. A good PFM solution should help them find the right answer. In fact, the scope of PFM is answering questions like these from a customer’s perspective.

There’s an abundance of misinformation, cookie cutter solutions, and armchair advisory products out there. In reality, the answers to these questions are unique to each customer. It all depends on the customer’s financial state, their financial behaviour, and future financial aspirations. In addition, macroeconomic conditions must also be taken into account. Answering through this lens is true personalisation in the financial domain.

Let’s get to the specifics of the solution in this scenario.

Every computer science engineer would know John von Neumann. He’s known as the father of the computer. Even modern day computers are built based on his basic principle of computer design: the von Neumann architecture.

To solve the problem at hand, one of the papers by John von Neumann comes to our rescue. The paper outlines a formal theory of expected utility theory, which talks about making decisions in risky situations.

While the expected utility theory is the right way to make decisions, no customer thinks that way. No human has the mental capacity to think that way on all financial decisions. This is where computers come in handy. They can find the right answer in a fraction of second. This is also where very creative Product Managers are required. The product construct is required to collect necessary information to address fundamental questions B1s mentioned earlier using the expected utility theory. And this is also where Behavioral Economics comes in to convince the user to take the right action to take.

You might’ve come across Daniel Kahneman’s bestseller, “Thinking, fast and slow”.

Expected utility theory talks about the right way to make decisions under risky situations. Kahneman and Tversky, in a paper titled Prospect Theory: An Analysis of Decision under Risk, talk about how people actually make decisions.

One of the key takeaways from the Prospect Theory is that people tend to value gains and pains differently.

Knowing how customers make inefficient decisions, one can frame nudges to help them make the right decisions. If you think this sounds unbelievable, check out the experiment conducted at Google to make employees eat healthy food.

To solve complex problems, we need the right scientific methodologies and product constructs that facilitate collecting data about customers’

• intent
• financial behaviour, and
• financial aspirations.

We need to build platforms like

• Audience Management to host knowledge graphs
• Communication Platforms to reach customers and nudge them
• Personalisation to help them take the right action
• Conversational UX to convince them of the right action in an empathetic conversational way.

We also need to build infrastructure like

• Data Platform to effectively manage customers’ data and democratise for solving customers’ problems with data.
• ML Platform to build, train, host ML models in a reproducible and iterative way.
• Experimentation Platform to increase the pace of development while not deviating from the scientific method.

Data Science in a for-profit setting is a team game. And to solve complex problems like these, we need people with diverse skills. At Jupiter, we’re solving these problems with a customer-first mindset, using Applied AI and Behavioural Economics.