The Commonwealth Bank wants to take banking “back to the 70’s”, a time when “the branch manager knew who you were” and customer trust often kept them with a single bank for life. That’s the vision of Dr Andrew McMullan, head of customer decisioning at CBA. He also believes AI is critical to achieving that goal.
During a presentation at Pegaworld in Las Vegas, McMullan shared how his team was using the AI-powered Pegasystems software to build customer engagement and trust.
CBA utilises Pegasystems’ “Next Best Action” feature – an AI powered CRM tool which provides contextual information in real time based on customer history and actions. According to McMullan, the bank has adapted the tool into its “Next Best Conversation” program where the information helps connect customers to the right staff who are armed with the contextual information and options.
The same technology also powers some of CBA’s automated digital features like chatbots and notifications.
But using AI in the bank’s digital channels is only possible now because of the measured and customer-centric approach taken from the start of the program, McMullan said, which began in branches and call centres, before eventually progressing to advanced use cases like chatbots.
At each stage the results are closely monitored and AI is only used if it creates a positive outcome for the customer, McMullan said. There are conversations where AI isn’t helpful and McMullan stressed each potential use case goes through stringent analysis and testing.
“If we ever feel that we’re not sure [about using AI], then we won’t build it,” he said.
However, McMullan is quick to caution, AI is no silver bullet and use cases must be appropriate, thoroughly tested, and always developed with customer outcomes and service in mind.
The strategy appears to be working, with CBA dominating the banking customer service rankings. Since claiming the top spot in banking customer satisfaction in 2013, Australia’s largest bank has fallen out of the lead in just nine of the following 59 months, according to McMullan.
No AI powered offers
Much of that success, McMullan says, comes from how the bank views its customer engagement opportunities. In the past customer engagement was, to its detriment, often seen as an opportunity to push offers to customers, McMullan said, noting most banks, including CBA, had moved away from the tactic.
Now the focus is on customer outcomes and how the customer’s view of the bank can be improved, according to McMullan.
“We genuinely believe if we just do more and more of those [positive engagements] the customer will just want to build a relationship with us,” McMullan said.
“We don’t incentivise our staff for selling or pushing any of the products, its just help the customer at the moment that they’re there trying to do whatever it is that they’re trying to do and build the relationship.”
He said the use of AI and Pegasystems was playing a large role in that mission, and the strategy required significant additions to their technology stack.
An AI stack
When a customer interacts with CBA they engage the bank’s “customer engagement engine” which determines the next best conversation available. This process, according to McMullan, occurs over 20 million times per day and results are returned in under 160 milliseconds.
At that scale it simply wouldn’t be possible using a traditional models based approach, McMullan said. So CBA developed an “AI stack”.
The first next best conversation models used in the initial branch phase were built manually, but CBA has been able to scale by employing Pegasystems adaptive model capability, where AI builds models based on data.
“We built 250 of the latest AI models in one week. My team of data scientists tell me that that would have taken them almost three years using their traditional modelling approach,” McMullan said.
CBA also uses natural language processing to automatically analyse customer responses for feedback, and AI to optimise testing.