How artificial intelligence can combat claims leakage

If insurers want to plug claims leakage once and for all, they need to introduce artificial intelligence (AI) — and the job shouldn't be reserved for the IT department, warns claims specialist Nigel Cade.

The Sydney-based managing director of ACDC Assessing and ANZIIF senior associate says the Australian insurance industry has fallen behind its Asian counterparts in embracing new technologies and an over-reliance on paper-based files and legacy systems is ultimately making claims leakage worse.

“The net result of this is the slow handling of claims, errors in settling of claims, processing of payments, the sale of inappropriate policies at the wrong price, an elevated pressure on customer service personnel, and a lack of control in identifying potentially fraudulent claims,” he says.

Claims leakage often peaks during a natural disaster response, Cade says, where inexperienced or undertrained staff are pushing claims through as fast as possible without asking the right questions or confirming important details.

“Nobody will give you a figure on claims leakage because they’re mostly in denial,” he says. “My view of leakage in terms of the claims I work on is that it’s very close to 30 per cent. But we shouldn’t call it leakage, we should call it missed opportunity.”

AI tools to fight claims leakage

Cade says AI’s potential for combating claims leakage and improving operational efficiencies is so broad that it’s helpful to think about it in four categories:

1. Risk and underwriting: “AI can be used by the underwriting department to be more thorough in profiling a risk at policy inception or renewal,” Cade says. “This will permit an insurer to price their product more accurately, taking into account specific needs to ensure that the customer only pays for the cover they need.”
2. Claims processing and operations: “AI can be used within the claims department to scan all documents, including completed claim forms, supporting tax invoices as proof of purchase, together with photographic evidence. Financial amounts being claimed can be processed through specific algorithms, meaning claims that pass the criteria can be immediately processed for payment to an insured. This will likely be most effective in high-volume, low-cost claims, such as motor, home building and home contents.”
3. Fraud handling: “Claims that fail the algorithmic process can then be reviewed by a senior person from the claims department, with a view to appointing a specialist loss adjuster to conduct further investigations, or refer the matter to the internal fraud department.”
4. Customer service: “AI can also be used within the customer service department, across a broad range of functions. Utilising chatbots, it is expected that front-line customer queries will be handled 24/7. This will deal with basic questions like ‘how to lodge a claim’, through to more complex matters like ‘getting an update on where my claim is at’.”

A company-wide cultural shift

Cade stresses that the move to introduce AI into business operations should not be left to the IT department to tackle alone. Rather, it should be a cultural shift across the entire organisation that must begin at board level. That’s because it’s 100 per cent dependent on ‘clean’ data, which Cade says is digital gold for insurers.

“[Many insurers’] data is still all over the place,” he says. “Without 100 per cent clean data, you’re going nowhere.”

Cade points to supermarket giant Woolworths as an example of a company that began cleaning its data more than two decades ago and is now streets ahead of where many insurers find themselves now. 

Cade helped to manage this process for Woolworths, beginning by digging deep into the company’s databases to look at just one product — blue swimmer crab.

“It’s sold as fresh, frozen and refrigerated, across multiple categories,” he says. “It was in the database 147 times, most times with different spelling.”

“That was a serious problem, and I had a team of 60 people to fix the company’s data. If that’s what your data looks like, you’re in real trouble.”

“I think for most Australian insurance businesses, motor data is over here, contents data is over there, and commercial data is in a different building. But they need to get it fixed, as overseas there are a lot of great initiatives going on and they’ll simply be left behind. In Asia and India it’s really happening. Some of the Asian insurers are going gangbusters.”

AI-driven systems fed by clean data enable customers to purchase, typically via an app, the type of flexible, customised insurance they expect.

“For example, a limousine driver might only want cover for a shift driving between 5am to 11am on certain days, while another driver wants cover from 4pm until midnight every day,” Cade says. “They can both get the gig insurance they need on the days they need it.”

The gift that keeps on giving

Data will very soon be the backbone of the insurance industry, Cade says. This data is collected daily, as local insurers conduct repetitive activities such as the selling of policies, collecting of premiums and paying out claims. 

“Data is currently regarded as static, and is used in the management information process to generate reports,” he says.

“The introduction of AI tools will significantly increase the number of decisions that can be processed. Within the claims department, human assessors can currently handle 25 files per day. Using AI tools, I would expect them to be able to process hundreds, if not thousands per day.”

This not only means greater efficiency, accuracy and much reduced claims leakage, but that the algorithm is learning day by day, becoming smarter, faster and more useful.

“It’s going to retrain itself as you go and get better with time,” Cade confirms. “It’s going to fix your claims leakage long term, but that’s just the beginning."

Find out more about fixing claims leakage using AI