AITech Interview with QBE insurance General Manager – David Bacon

QBE Insurance General Manager David Bacon Addresses Challenges and Opportunities as AI Reshapes the Commercial Insurance Industry

1. Tell us a little bit about QBE and the key issues that needed to be resolved or addressed.

QBE has been around since 1886 and has built a global presence. Our Australian business was actually the first to apply machine learning and artificial intelligence to personal injury claims within QBE.

About three years ago, we were running what I would call a process-compliant business, like most other insurers. We liked the idea of having more complex claims managed by our best people, but in order to achieve that, we were doing a lot of manual triaging of claims. Our internal experts essentially reviewed claims right after they were logged in order to identify potential risks in the portfolio. This was based primarily on whether the person who logged the claim was losing time from work. We knew there had to be a better way.

The second aspect that we were having trouble with was basic models of case load each adjuster was carrying — it wasn’t well aligned to the risk in the claims. This is not unique to QBE, but an adjuster might be assigned 70-80 claims at any one time, and with this volume, it is next to impossible to know how complex the claims are within that mix.

Another issue we wanted to solve related to our quality processes. At that time, team leaders were manually selecting open files for a brief quality assessment. It was a bit of a guessing game, so there was a low level of confidence that we were selecting the claims that would benefit most from a quality review.

2. What were you looking for in a solution to help with these issues?

Initially, we considered building a solution ourselves but realized it would take a tremendous amount of staff time and resources. I was getting pressure internally to work with one of the more historical data and analytics (D&A) shops. Most carriers at that time were reaching out to D&A consultancies, so we engaged with them. Unfortunately, most of what they showed was that they were interested in billing day rate work. Moreover, they seemed to end up, at times, solving a very narrow slice of a problem that tended not to be scalable or actionable.

We believed that along with deep knowledge of machine learning, whoever we chose to develop a solution should be trying to solve the same problems we had identified.

It was important to us that the vendor was quite well aligned with our philosophy that the right claim should be directed to the right adjuster. CLARA Analytics was a fit for us in this regard.

3. Why did you choose CLARA Analytics? What made CLARA stand out, or what feature of our product was most appealing?

CLARA’s technology could identify the risk in a claim appropriately, quickly and accurately so that a claim could be managed by the best person for that claim. As soon as we examined how CLARA’s models worked and how the software visualizes workload allocation, we recognized that it was exactly the way we wanted to run our business.

Accuracy was of the utmost importance to us. By adopting CLARA, we were able to stand up an AI-based solution much faster than we could develop our own internal model, and CLARA’s work would give us a head start with regards to accuracy within the model. In addition, their user interface is quite elegant. It consciously strips out a lot of the complexity while providing a handful of actionable signals, which is what we need in our business.

While these things are great, what it ultimately came down to was trust. After spending time with the team, it was clear that we were very like-minded. I knew that I could learn a lot from the CLARA team and hopefully find ways to contribute to making AI-based approaches more widely accepted in the insurance industry.

4. What measurable benefits have you seen? Has this solution saved money and/or increased productivity?

There are a number of benefits worth highlighting. We derive financial benefit out of our application of CLARA’s models. The initial reports are indicating that we should easily see a 5:1 return on investment. That is a conservative estimate, and that benefit number is only on half the portfolio (workers’ comp) right now.

With CLARA, we are able to implement a more focused approach to quality assurance. Instead of the leadership team selecting claims for quality review, we now follow the system’s signals so that any escalations in complexity are flagged for quality review. That gives us a much higher level of assurance that we’re reviewing the most appropriate claims. It is our belief that quality assurance shouldn’t be driven by art; it should be driven by data science, which is exactly what we’ve been able to accomplish.

Additionally, there has been a substantial competitive benefit. The efficiency and insights we can achieve now have become a very strong part of our commentary in terms of what we’re selling. Our customers are very interested in this machine learning, bleeding-edge technology. It gives us a real sales advantage.

Another clear benefit stems out of how we collaborate with the CLARA team. For example, over the last several months, we’ve been collecting perception data from each injured person’s claim, such as how they are feeling about their recovery. Today, we’re able to collect and analyze that information at scale. This is a big advancement. People have been talking about psychosocial flags for injury recovery for more than 20 years, and no one has solved it until now.

There are also important intangible benefits, such as the team benefit. Our employees love receiving information that allows them to become more effective and that makes their jobs easier. What we realized is that once you start to peel away some of the change management and people’s initial reluctance to AI, they actually come to understand what the system is doing, and then they’re motivated by it. We can now strategically assign adjusters to claims that align with their skills, which is a nice thing for both the adjuster and for us operationally.

5. What have you been most impressed with so far?

I think at a high level, how easy it’s been to implement an AI-based solution that tells us what we need to know so that we can apply the information in a useful way. This helps our claims process, enhances the quality of our offering, and allows us to serve our customers more effectively.

One thing that has surprised me is how willing CLARA’s team has been to consider other approaches and think about problems in a different way based on our discussions. I find this to be quite rare as CLARA essentially operates an off-the-shelf model. Other vendors typically shut these types of conversations down, or they go to the other end of the spectrum where you wind up solving some esoteric problem that was never really a problem. From the buy side, I value this level of engagement as well as the fact that CLARA understands problems/solutions in addition to its product well enough to adapt to our needs.

6. Is there anything else we should know or that you want to say?

Good AI models can be transformative. In our case, CLARA and the information its products provide have become crucial to our operations. The insights we have gained truly change the game in how we handle workers’ comp claims. We’ve been so pleased with the outcomes that we are now using CLARA for auto liability as well.

David Bacon is general manager at QBE Insurance, one of the world’s top 20 general insurance and reinsurance companies, with operations in all the key insurance markets. QBE is listed on the Australian Securities Exchange and is headquartered in Sydney. For more information, visit www.qbe.com. For more information on CLARA Analytics, the leading provider of artificial intelligence (AI) technology in the commercial insurance industry, visit www.claraanalytics.com, and follow the company on LinkedIn and Twitter.