How Toronto’s integrate.ai aims to turn corporate data into profit

Open this photo in gallery:

Steve Irvine has made the bottom line argument for large corporations to adopt AI since he moved to Toronto from Silicon Valley to start integrate.ai Inc. in early 2017.

The pragmatic 39-year-old’s goal was “to bring the type of toolkit you’d more commonly find – and that has been purpose-built – inside an Amazon or a Facebook or a Google into the large enterprise.” He wanted to build software easy enough for a non-technical user that would combine a number of AI techniques – including deep learning and Bayesian models, which determine probabilities based on past experiences – to help companies make predictions by mining their data.

Mr. Irvine raised US$10-million in venture capital, recruited star employees such as New York AI consultant Kathryn Hume and attracted advisers such as University of Toronto computer science professor Richard Zemel, who heads research at Toronto’s Vector Institute for Artificial Intelligence.

But Mr. Irvine also knew he needed a story to tell the market. So he set out to work with five “alpha” corporate customers to deploy AI on a few projects, use their feedback to influence integrate.ai’s software development and share their experiences.

As AI goes corporate, startups look to prove themselves

In addition to Corus Entertainment, the company signed Bank of Nova Scotia and online insurance quote provider Kanetix Ltd., plus an undisclosed auto manufacturer and gaming firm.

Integrate.ai’s early goal was to help companies allocate marketing dollars more effectively by not pitching to those most or least likely to buy and instead focusing on the middle 70 per cent. ”You want to go for those people where, if you were to do something differently, you could modify their behaviour slightly to get a bit more return,” Ms. Hume said.

Scotiabank started integrate.ai on a project with Cencosud, its Chilean credit card business. Cencosud had been blasting e-mails to potential customers and wanted to know if it could drive up such key marketing measures as open rates and customer satisfaction by more selectively targeting those likeliest to bite.

“If you can drive a couple of efficiency points here and there, you’re saving a tremendous amount of money … and hopefully driving revenues,” said Scotiabank chief digital officer Shawn Rose.

By combining data sets and training its algorithms to look for patterns, integrate.ai helped the bank home in on the right customers. In one campaign last year, Cencosud bumped up revenues by 40 per cent over target, with projected financial benefits exceeding C$1.5 million. Now the bank has asked integrate.ai to help its Canadian retail banking business. “If we can apply this thinking to a bunch of different areas at the bank, it’s going to make this really efficient and make us some money,” Mr. Rose said.

Kanetix CEO Andrew Lo was similarly looking to increase the number of visitors to his website who would go on to purchase policies. Kanetix had lots of data: Car insurance shoppers typically answer 50 questions on its website, and purchases generate more information. (Mr. Lo said Kanetix ensured that any data shared did not identify customers individually.)

Integrate.ai’s algorithms drew correlations from various signals in the data to target the best prospects. Kanetix already knew from past surveys that visitors were likelier to buy insurance if they could save more than $500 over their existing policies. But integrate.ai was able to identify other signals of purchase intentions, including postal codes and the type of car a visitor drove (a Honda Civic driver in the suburbs, for example, would be more easily swayed by a $500 savings than a Porsche driver in Rosedale or Westmount), as well as gender and number of years insured.

By stratifying visitors according to their likelihood to buy, Kanetix could target individuals with different landing pages – the web pages where people end up when they click on an ad – while selectively offering $20 gift cards to those who needed more of a nudge. In a business where effective campaigns might increase conversions by one percentage point, integrate.ai’s insights led to a 13-per-cent bump for Kanetix. Mr. Lo had initially expected no improvement. Now, “we’re going to start scaling [the use of integrate.ai ’s tools] within our organization,” he said.