Speaking to The Canadian Club of Montreal on October 28, Isabelle Girard, Senior Vice-President and Chief Data and Digital Officer of Intact Financial Corporation, outlined the company's efforts to use generative artificial intelligence (AI).
Girard, an actuary by training, was one of three guests invited to this panel discussion focused on artificial intelligence. Since January 2024, she has been head of Intact Lab, which employs around 500 people. Since 2017, this lab has included a data science component, which helps the insurer make the most of the large volume of data that is collected on policyholders, risks and claims.
“We see artificial intelligence as a tool to give us new approaches to better leverage all the data we need within our business,” she says.
According to Girard, the approach is very practical. “Our aim is to use artificial intelligence to deploy solutions within the various units. Since the creation of Intact Lab, we've deployed around 370 models in the company’s different sectors. So, what works for us is our centralized approach, developing expertise in-house and then working closely with the business units to understand how to apply artificial intelligence solutions within their reality.”
Pricing
The first applications of AI were used “to better segment our premiums and in our insurance pricing. We also use AI to manage a large amount of data in our telematics program,” says Girard. For example, we process customer driving data for car insurance.
“We also use artificial intelligence solutions to speed up data entry in our commercial insurance segment. The system retrieves information from different types of documents, be they e-mails, Excel or PDF files, and then we use artificial intelligence to enter this data into our systems so that we can generate quotes more quickly for our customers,” she explains.
A centralized system
She recommends that companies wishing to take advantage of generative AI adopt a centralized approach “to ensure we develop expertise, but also learn from mistakes as we deploy models.”
The insurer learned to develop its models by applying them to an area in which it has expertise: risk segmentation. After a few years, deployment continued in other areas.
“In our experience, you need to ensure that your management team is a good partner for artificial intelligence projects. They also need to understand what artificial intelligence can do, but also understand its limitations, which is just as important,” she says.
Girard adds that it's important not to underestimate the time required to adequately prepare the data that will be entrusted to the AI platform. “An AI model without data doesn't yield much. We hadn't necessarily assessed all the time and energy it takes to prepare the data, make the data accessible and then deploy the AI and make models.”
She points out that not all models have to be developed in-house. The insurer also uses some of these tools in open source mode. According to her, the important thing is to make them available to employees in a secure environment.
The tools developed using AI aim to improve employee productivity so that they can focus on tasks that “bring value”.
“We're using AI to speed up data entry, so the underwriter can spend more time analyzing risk, instead of entering information from one system to another,” says Girard.
The enhanced underwriting platform makes it possible to target elements and draw the underwriter's attention to an aspect of the risk they are analyzing. “It's really important that the human element is part of the process,” she says.
Staff training is required, not least to challenge the accuracy of the answers provided by generative AI. “It's not perfect,” stresses Isabelle Girard. The metropolitan region (of Montreal) benefits from a highly developed ecosystem when it comes to AI-related training, and companies would do well to take advantage of this, she stated.
She encouraged participants to get started using generative AI. “You will find your way to bring additional value to your business with AI,” she concludes.
