Canadian Underwriter

Building intelligent distribution

July 18, 2018   by Eric Chalmers

Print this page


Artificial intelligence (AI) technology has been called “the fourth industrial revolution in insurance,” and the global insurance industry is excited about it.

A TCS Global Trend Study released in 2017 suggests that the average major insurance company spent $124 million on AI in 2015—much more than the cross-industry average of $70 million. The same study predicts spending to slowly decrease to $90 million by 2020, suggesting companies are already transitioning from R&D to deployment.

At Surex Direct, we’re doing our own AI and machine learning R&D in-house. We’ve deployed several intelligent systems, and have others in various stages of development.

There’s a lot of talk in the media about AI putting people out of work, but we don’t think that’s a big concern in our business. We believe AI works best when used to enhance human decision-making, rather than replace it. It’s a tool that will allow us to extend our reach and let us offer a level of service that all brokers want to offer.

But with so much excitement around AI, it can be hard to separate inflated expectations from realistic ones. The following may be useful to brokers interested in learning more about the value of AI.

Using AI to automate and optimize

A new billing tool we developed has allowed us to reclaim three to six hours a day in productivity. It uses AI to automatically read and categorize billing documents, perform data entry and lookup, and notify customers when there is a billing issue. These tasks used to be performed manually, consuming time that is now spent on more meaningful customer interactions.

Our billing document handler came online in November of last year, and its performance slowly improves as it learns. It now handles about half of all billing documents autonomously, including initial contact with the customer in most cases.

AI is well suited to automating routine manual processes, which not only improves efficiency but also the satisfaction of staff who are relieved of the (usually tedious) tasks.

Another project underway at Surex is an intelligent auditor: a machine learning–based tool that reviews all policy changes and transactions in an effort to catch mistakes. By analyzing data from these transactions, the auditor will learn, for example, to identify coverages and endorsements that are normal for a particular situation. So, if the broker forgets to request a limited waiver of depreciation endorsement for a new vehicle, the auditor will flag that case as unusual and refer it to staff for review.

of digital transformation initiatives will use AI services by 2019

Source: International Data Corporation

US $88.3 million

The global chatbot total market value in 2015

Source: Credence Research


of respondents to a recent study believe AI will help create jobs, rather than destroy them

Source: Capgemini

We have audit staff who do this job now, and they are great at it. The problem is that mistakes are rare. Most cases are considered normal, so staff spend a lot of time reviewing non-issues. We want AI to take the bulk of the easy cases off their plate, so they can focus on the exceptions.

Dreaming big: Creating new services using AI

Chatbots are an AI technology forecasted to play a major role in the industry, for reasons that are obvious to anyone who has ever been stuck on hold.

A good chatbot can replace a clunky user interface, or automate a predictable customer interaction. But AI opportunities go beyond chatbots—they can enable completely new services, some of which we’re working on right now.

One example is a new service codenamed “Pulse,” which will automatically and intelligently monitor the market on each customer’s behalf, like a personal insurance agent.

Another example will apply machine learning techniques to recognize patterns in Surex’s unique database, effectively allowing Surex to perform a type of actuarial science. We don’t have more data than insurance carriers, but we have broader data. We don’t get to see carrier’s risk prediction models, but we do see hundreds of quotes a day, across multiple carriers. And each quote reveals something about how that carrier models that risk. We also handle claims across multiple carriers. Insights into that data could be valuable to the industry.

Thanks to AI and machine learning, we may begin to see changes in how insurance pricing and actuarial processes are performed, with brokers playing a bigger role than ever before.

“We believe AI works best when used to enhance human decisionmaking, rather than replace it.”

Positioning AI within a bigger technology strategy

Nearly all Canadian brokerages have digitized to some degree, and some have developed custom software. But there’s a key difference between AI and conventional software.

Software needs to be updated constantly to reflect changes in the industry. An intelligent system, on the other hand, has some ability to adapt to change. Our systems learn by observing staff members’ actions, or from other data sources, but the central feature and goal is always the ability to self-improve with little intervention from the designer. AI complements conventional software and should have a place in any technology strategy.

There are three primary methods of procuring AI technology that brokers can consider:

  • Engaging contractors for specific projects: Data science consulting firms can provide instant expertise. Costs are high, but this may be the best option for a broker who wants to deploy one or two specific AI systems, especially if the contractor has developed similar systems before. Contractors should be selected with care, and a method should be established for the broker (not the contractor) to measure success.
  • Developing systems in-house: We consider AI development to be a long-term, transformative process, making permanent data science staff more cost-effective than contractors. In-house data scientists also learn the industry better than a contractor would, enabling them to see opportunities for services that a generalist contractor might overlook.
  • Licensing industry-specific technologies: A forecasted shortage of data scientists could make in-house expertise difficult to find. A third option is to licence finished products that are ready to be deployed. Some of the AI components we’ve developed could be used by many brokers in Canada, and we’re open to licensing deals that would help move the industry forward.

Eric Chalmers is the Director of Data Science with The online brokerage launched in 2012 and is active in six provinces and two territories. Still headquartered in Magrath, a rural Alberta town with a populationof 2,300, developing and implementing technology is essential to Surex Direct’s day-to-day business.

Copyright © 2018 Transcontinental Media G.P. This article first appeared in the April edition of Canadian Insurance Top Broker magazine

This story was originally published by Canadian Insurance Top Broker.