After paying a nine-figure sum for some of the brightest minds in Canadian artificial intelligence, Toronto-Dominion Bank is trying to use tech to bring back what tech has greatly diminished: the personal touch of a trusty branch manager.
With its acquisition of Toronto predictive-artificial intelligence (AI) company Layer 6 last year, TD is aiming to better foresee client needs and show up at the right time with the right products and services. What was a startling acquisition at the time – the price was not disclosed, but The Globe reported that TD paid more than US$100-million-plus for the barely year-old company with just 17 employees – is starting to pay off, the bank says.
TD chief AI officer Tomi Poutanen, a co-founder of both Layer 6 and Toronto’s renowned Vector Institute for Artificial Intelligence, uses the example of prospective home buyers as clients whose needs his now-34-person team is trying to predict.
“We don’t want to be there only on the day that they’re shopping around for the mortgage,” Mr. Poutanen said in a recent interview at Layer 6′s office at Toronto’s MaRS Discovery District.
That doesn’t mean just making mortgage offers – it means recognizing the signals that preface home buying, such as when a client’s income nears the point when they can support a mortgage. “If you know far enough in the future … you may start giving them savings instruments well before you’re talking about a mortgage.”
While the bank is hesitant to share exactly how much more efficient Layer 6’s AI has made TD at predicting mortgage needs, Mr. Poutanen says that across six machine-learning projects his team has run with TD data, it has improved the accuracy of advice and product offers between 30 and 57 per cent. “It completely changes the dynamics of the business," he says.
Canadian banks have been embracing AI in recent years with various goals, including cutting costs by doing some of the tasks humans used to perform and keeping pace with financial technology startups hoping to disrupt their businesses. Some uses are customer facing, such as the Bank of Montreal’s AI-powered “chatbots” to guide clients online. Royal Bank of Canada’s latest mobile app, meanwhile, uses AI to scan paper bills and pull out the necessary details to make payments simpler.
Other efforts are more behind the scenes, such as RBC’s funding of research through subsidiary Borealis AI. National Bank of Canada is collaborating with Montreal’s Element AI on cybersecurity, and has been embedding AI experts across the bank. The Bank of Nova Scotia teamed up with Toronto AI startup Dessa in 2017 to analyze credit card repayments, and also works with Toronto’s integrate.ai, whose software helps businesses predict customer needs to increase retention.
In recent years, integrate.ai CEO Steve Irvine has watched financial services companies embrace AI in two broad ways. One is automating mundane and low-value tasks such as data entry to save human time for human judgment. The other is delivering more nuance to decision-making by analyzing data from tens of millions of customers to spot sales and retention opportunities, similar to what Layer 6 does.
“If I can do even a slightly better job in those situations – of making a better probabilistic guess – then they’re massive funnels with huge revenue and profit implications,” Mr. Irvine says. That could be through offering services at the right time, managing trades more effectively, or changing credit models – the sky is the limit. “If we make decisions slightly better, that’s billions of dollars."
TD, as it happens, is attempting to use Layer 6’s machine learning to overhaul its approach to customer service. Before the Internet, “the branch manager knew everything about a family’s financial needs,” said Michael Rhodes, TD’s head of innovation, technology and shared services. But between call centres, apps and ATMs, today, “the insights we have about customers are distributed throughout the organization. … It’s very difficult to have the full context of a customer relationship the way you could have 40 years ago.”
The TD-Layer 6 partnership emerged at a time when both parties were facing transitions. Founded by Mr. Poutanen, entertainment lawyer Jordan Jacobs and AI PhD Maksims Volkovs in late 2016, the AI company had been spun out from Milq, a startup that recommends personalized cultural content.
Its team drew global attention almost immediately when, in 2017, it beat out numerous Big Tech companies to win RecSys Challenge, a renowned AI-recommendation competition. Soon, Mr. Poutanen says, the company was at a crossroads, as it considered deals from venture capitalists who were more focused on sales than engineering – while also drawing the attention of a Big Tech company he won’t name that made an acquisition offer at a valuation that would have pleased shareholders.
As tempting as that was, he says, the Layer 6 team prided itself on hiring and supporting homegrown AI talent. In swung TD, and a chance to stay and hire in Canada. The bank had been centralizing data from its 25 million clients into what it calls a “data lake,” making it ripe for analysis with the kind of AI Layer 6 had developed.
For the past year, Mr. Poutanen’s team has run analyses and experiments with TD’s data lake, tracing patterns across all of its business lines to determine what might predict something else. “The work before [the acquisition] and today is similar in nature, but at a much greater depth now with TD,” Mr. Poutanen says.