NAB says 800 adaptive models drive its 'customer brain'

NAB is running around 1000 data attributes about its customers through 800 adaptive machine learning models at any one time to power its ‘customer brain’ next-best action system.

NAB's Jess Cuthbertson.

Customer analytics and decisioning executive Jess Cuthbertson showcased the customer brain in a keynote at the PegaWorld iNspire 2024 conference in Las Vegas this morning.

The personalisation engine was flagged as being under development back in 2022, with further details of the project published in The Australian newspaper earlier this year.

The customer brain is underpinned by Pega’s customer decision hub platform.

Cuthbertson said that in line with the bank’s cloud-native approach to its systems, NAB’s Pega environments were “up and running within three weeks of signing our contract.”

The system has been live for around two years and covers three-quarters of customer interactions.

“For our customer brain program, we went from zero to 75 percent coverage of customer interactions in less than 24 months,” she said. 

“We delivered over 150 next best actions across service, sales and engagement type experiences.”

Those actions ranged from in-app payment reminders, to messaging designed to keep customers informed at each stage of the home loan process, and “milestone” celebrations, such as “meeting a savings goal or paying down your mortgage”.

Cuthbertson said that 65 percent of the bank’s customers on average “hit the mobile [banking] app at least twice a day.”

“So, we’ve got a great opportunity to engage with them, connect with them and just keep a conversation going,” she said.

The bank spent considerable time upfront both “listening to customers through data” and understanding from frontline bankers how their experience told them that a customer might be having problems.

“We really took our time upfront with the customer brain program to listen to the customers through data - what are they telling us? We looked at net promoter score, at complaints data, at different journeys where you can see customers were getting stuck along the way, and we used that to prioritise our experiences,” Cuthbertson said.

“We also sat with our bankers [to understand] what do they look for when speaking to a customer, how do they see the customer might need a little bit of help, what does that look like? So, we took that, we codified it, and put it into the system.”

Cuthbertson said that the machine learning algorithms sitting behind the customer brain also monitored the efficacy of the action, and how customers engaged with it.

“At the moment, I think we’ve got about 1000 data attributes feeding the customer brain - just different things that we can see in customer data - and then we’ve got probably 800 adaptive models running at any point in time,” she said.

“So, we’re taking the data we have about the customers but then we’re also listening and hope they’re interacting and responding in real-time to fuel the engine.”

Cuthbertson said the bank experienced a 40 percent “lift” in customer engagement “as soon as we plugged the brain in”.

“I think it shows the fact that we’re being more relevant,” she said.

“It might mean that we actually speak to some of our customers less but when we do, they can trust it’ll be with something that’s relevant and in the moment for their needs.”

The customer brain has also given bankers more time and confidence that conversations with customers will be well-received.

“Bankers can spend time with customers, which is what they want to do,” Cuthbertson said.

“Whether you’ve been a banker for five minutes or five years, it can help create consistency and confidence to be able to pick up the phone to customers.”