Inside NAB’s customer brain and how it’s taking banking ‘back to the 70s’

Using billions of data points, NAB has built what its calls its ‘customer brain’, which aims to solve banking problems before they arise.

NAB says its “customer brain” — built from billions of data points — aims to catapult banking back to the 1970s, a statement which appears at odds with the high-tech project.

But, before you think the top four bank is diving into a cheesy pot of fondue, think again. Chief data and analytics officer Christian Nelissen says the retro nature of the project is aimed at making the technology work as “naturally” as possible.

For example, the “brain” sends happy birthday messages, something which isn’t entirely new in customer relations, but is a way to increase engagement and make the technology appear less distant.

NAB chief data and analytics officer Christian Nelissen.

“I sometimes talk about this as taking banking back to the ’70s,” Mr Nelissen said.

“In the ‘70s when you walked into the branch once every two weeks, do your pay or whatever and the people in the bank knew you and it felt natural that someone wished you a happy birthday.  

“We’re all people, right? We don’t see people now. Human to human contact is going down. But being able to replicate that very natural experience, with these engagement messages, is important.”

The project, years in the making, effectively uses AI to analyse customer habits and break down silos which can build up in large organisations to get a better view of the customer and anticipate their needs.

It is powered by 200 machine-learning models and has so far taken in 36 billion data points, covering everything from transactions to the way customers use the NAB banking app.

The “brain” is designed to deliver the precise products to the right customers while sorting out day-to-day banking problems before they arise. Crucially, it’s about driving customer loyalty and stopping leakage to rivals.

Aside from sending happy birthday messages it can solve challenging banking problems. Nelissen recalls a story about a Barclays bank manager calling a customer into a branch in the 1970s. On arrival, the manager asked for the customer’s credit card and cut it up.

“He said ‘you’re not using it properly. You’re not paying it down every month, you’re carrying debt on it. It’s the wrong way to use this card’,” Mr Nelissen said.

“I love that story because it connects so strongly with the way we’re thinking about what we want to do. The bank manager wasn’t thinking ‘I’m going to make more interest’. The bank manager was thinking ‘that guy is going to get into trouble unless I give them a prompt here’.

“I don’t know whether you could cut someone’s credit card today and not get a complaint from the regulator, but the idea is actually quite interesting.”

A more modern take, using the power of the big data “brain”, is asking a customer whether they want to keep a higher daily withdrawal limit after increasing it for a one-off purchase.

“We have messages that go out to those customers with higher limits to say, ‘hey, have a think about do you need that higher limit’,” Mr Nelissen said.

“About 10 per cent of the customers go in and take the limit back down. So that’s a positive service message and you can see how that connects with the data that we have about the customer.”

Before using machine learning and drawing on the power of big data, such a task would have taken a human a considerable amount of time, particularly when the personal relationship with local branch staff is becoming a fading memory.

“Until the brain came along, those things you could always do,” Mr Nelissen said. “You could do any one thing. But you couldn’t do a lot of things. And what the brain gives us is the ability to scale those messages. It becomes economic to have a conversation with a relatively small group of customers that is meaningful to that small group.

“Before we’d have to go too large, because of the cost of the effort and our internal friction cost of doing things, so you couldn’t always be as relevant. Here we can have much smaller groups and be much more relevant.”

The banks have been among the earliest adopters of machine learning and artificial intelligence, with German bank Commerzbank even launching an AI-powered “avatar” — what it describes as a “digitalised person” — last month to perform various banking services, including personalised advice.

Commonwealth Bank is planning to follow suit but says it will be a gradual process — first providing information only, before eventually facilitating transactions. Mr Nelissen said NAB had a similar strategy.

“The first step is to have the bankers use the tool, so when they get a question from the customer … they intermediate some of what you get back to make sure that it’s right. We’re learning about how it responds and what we need to do to make it respond in a better or the right way.

“We’ve got chatbots and stuff like that and some of those chatbots work in a more deterministic way. We were more comfortable with those.

“But the ability to do generative AI stuff externally, that’s on our horizons, but when will we get there? It won’t be years. But it won’t be weeks either. So it’ll be somewhere in between when customers start to see unfiltered views or what’s coming out of that environment.”