Skip to main content

The Globe and Mail

Scotiabank teams up with Toronto AI startup DeepLearni.ng to collect debts

DeepLearni.ng’s Stephen Piron, left, and University of Waterloo co-op student Tristan Monger at the company’s Toronto office.

DeepLearni.ng

When Bank of Nova Scotia customers miss a credit card payment, an artificial intelligence system is shaping how the bank tries to get its money.

The system, which was completed in December, is the result of a partnership between the bank and DeepLearni.ng, a Toronto-based AI startup.

It analyzes the individual customer's past behaviour, as well as the bank's large data set around credit card collections, looking for connections.

Story continues below advertisement

"You're trying to find the patterns of behaviour that lead you to believe a customer will pay you back," says Neil Bartlett, Scotiabank's senior vice-president of Analytics.

Overdue payments are "something that all banks have been dealing with as long as credit cards have existed," says Michael Zerbs, executive vice-president and co-head of Information Technology and Enterprise Technology at Scotiabank. "The problem that we all have is it's actually not that straightforward."

The big challenge is finding the right "treatment," Mr. Zerbs says, and getting different customers to pay takes different measures.

For example, he says, if a customer is a few days late and the data suggests they will pay on their own, no action may be necessary.

In other cases, the data could recommend calls or e-mails encouraging the customer to come to a branch or bringing in a collection agency.

The system developed by DeepLearni.ng uses a type of artificial intelligence called deep learning, also known as a neural network. It's modelled on the human brain and combines multiple layers of processing.

Stephen Piron, the founder of DeepLearni.ng, says that while his company has worked with Scotiabank in the past to do research, the current collaboration began over the summer.

Story continues below advertisement

"They're keen to use AI in different parts of the bank. They wanted a good-use case that was easy to measure and that had some economic value," Mr. Piron says.

For several months, the AI system was tested against the bank's existing model, using historical data, to see which would generate better results, Mr. Piron says. His technology came out ahead.

"Then it was a mad rush in the last few months of 2016 to put it into production," he says. "Now we're in a third phase where it's about incorporating more data sets, understanding the feedback loop about the treatment, the way we interact with credit card customers, then potentially broadening out to things that look like credit cards; so car loans, mortgages, small-business loans."

One of the biggest challenges was making the AI work with the bank's systems.

"If you're clever, you can probably build a model that will beat an existing model, but actually making it work, with all of the nuances of a live credit card system for millions of people, that's a really big challenge," he says.

The partnership sees Scotiabank paying DeepLearni.ng for consulting, as well as licensing its software. Mr. Piron declined to comment on the value of the deal.

Story continues below advertisement

Joel So, a partner in PwC Canada's Financial Services Consulting practice, says that while AI technology has reached a point where it can be commercialized, banks are still trying to figure out the best ways to use it.

"Right now, the partnering model works well because it allows the existing startups and research centres to continue doing what they're doing, but with a level of influence and collaboration with the large banks," he says.

"For a bank to be able to get close to those technologies now and influence what types of technologies become commercialized later for their benefit, that's really what they're trying to go after," he says. "The question then becomes, is it more effective to help shape this stuff by doing it in-house, when you didn't have that capability at all, or to plug into some thriving ecosystem?"

Scotiabank is also using a personalization engine developed by Toronto's Layer 6 AI to help market its products to people who have the Scene loyalty card, itself a joint venture between Scotiabank and Cineplex.

Jordan Jacobs, the co-CEO of Layer 6, says it allows the bank to go beyond the mass marketing approach it's taken in the past.

"Their thinking is, 'Can we personalize what we offer to people, based on our understanding of them, and offer them the kind of thing that would make them happy,'" he says.

That could mean offering some sort of benefit in the loyalty program, or a targeted marketing message.

"It's a win-win because we spend our marketing dollars more wisely and customers get things that make sense for them," Scotiabank's Mr. Zerbs says.

Other banks are also interested in AI. Mr. Piron says he's had talks with all of Canada's major banks.

And there's a good reason, Mr. Bartlett says.

"The great thing about deep learning as a strategy and a technique is you don't have to have figured out all this up-front. The data can actually tell you the right thing to do," he says.

Report an error
Comments

The Globe invites you to share your views. Please stay on topic and be respectful to everyone. For more information on our commenting policies and how our community-based moderation works, please read our Community Guidelines and our Terms and Conditions.

We’ve made some technical updates to our commenting software. If you are experiencing any issues posting comments, simply log out and log back in.

Discussion loading… ✨