Insurance brokerages certainly recognize the value of cognitive computing, but they’re taking a gradual approach to its use. The experimentation makes sense, proving the value of the tools and serving as a springboard for their wider future use.

“The toughest part in deploying new technologies is making the business case for them,” says Stewart Gibson, senior vice president and chief information officer at USI Insurance Services, which recently agreed to purchase Wells Fargo Insurance Services USA. “Early adopters bear the bulk of the cost of a new technology’s implementation. We’re more interested in being a fast follower, tiptoeing into these new solutions as their value becomes clearer.”

Here’s a brief look at some of the new technologies being deployed at a few large brokerages.

“RPA is probably the most prevalent of the new cognitive tools we’ve been using,” Gibson says. “We’ve deployed an RPA tool from Cisco to perceive threats within our network infrastructure.”

Like other large brokerages, USI’s network overflows with millions of transactions daily. Any one of these interactions can cause havoc if the data being transmitted are infected with malware.

“We’d need an army of network engineers watching all the end points inside the network to discern trouble spots,” Gibson says. “The RPA tool provides this vigilance with little manpower. It’s smart enough to identify what might happen or something that is about to happen right now. Network traffic is then automatically rerouted over backup circuits to limit the potential downtime.”

The brokerage also has implemented a matching tool using machine learning to assist salespeople in configuring risk transfer solutions covering customer risks in the middle market. Called Omni, the tool draws insights from USI’s comprehensive database containing risk-based information on more than 100,000 clients over the past 100 years.

“A rules-based engine populated with algorithms, decision trees and linear regressions helps pinpoint the specific insurance coverages and financial limits of protection a client may need,” Gibson says. “It analyzes clients in a similar industry that had chosen to cover their risks in a specific way and then presents the financial outcomes of these decisions—did the insurance fully protect them or just partly protect them? It then configures a suite of customized insurance solutions for the client.”

USI built Omni from scratch using spreadsheets six years ago. The tool has evolved in the intervening years and has recently been turned into a web application. “All it takes is one button to generate a client-ready presentation,” Gibson says. “Since this is a machine-learning solution, each time we bind new coverages for a client, the tool incorporates this new knowledge in future presentations.”

USI is just starting to evaluate artificial intelligence solutions using natural language processing. “We like the idea of processing simple customer requests without having to be on the phone,” Gibson says. “And we also like solutions that pick up inferences from a customer’s words to fix a problem without having a human involved. We may explore a proof-of-concept of these self-service tools down the line but are not yet ready to go mainstream.”

Like other brokerages, Hub International keeps a vigilant eye on developments in the bustling insurtech sector, where hundreds of innovative startups are developing novel cognitive computing solutions for the insurance industry.

“They’re constantly coming up with new ideas that are challenging the status quo in the (insurance) ecosystem, forcing us all to think about doing things in more efficient, cost-effective and customer-centric ways,” explains Carla Moradi, Hub executive vice president of operations and technology.

One way Hub is utilizing machine learning is to compare its insurance policy submissions on behalf of clients with the actual policies that are issued. Brokers typically rely on people to perform such comparisons, which absorbs time and effort that can better be devoted to customer needs. “The process now takes seconds,” Moradi says. “If an error is identified, humans take it from there.”

Hub also is studying the use of RPA to further streamline workflows. “We haven’t deployed anything yet, but we’ve identified several key areas where there are repeatable processes involving the same keystrokes,” Moradi says. “RPA can be used to do these keystrokes, increasing the amount of time our people spend with customers.”

Down the line, Moradi predicts that brokerages will collect and analyze risk-based information generated by their clients’ gradual embrace of the internet of things. “This is yet a great opportunity for us to provide another high-value client service,” she says.

Few lines of insurance are more data intensive than workers compensation, given the voluminous data produced by hospitals, physicians, nurse case managers, pharmacies, physical therapists, insurance carriers, claims adjusters, and other parties engaged in returning an injured employee to work. Using data mining, natural language processing and machine learning, Lockton is accessing and analyzing workers compensation claims to improve the employee’s experience.

“One of the most common cost drivers of workers compensation is a lack of communication with the claimant, which increases their fear and the possibility that they may hire a workers compensation attorney,” says Mark Moitoso, Lockton’s senior vice president and analytics practice leader. “To better understand how a claimant is feeling, we’re relying on text mining and machine-learning technology.”

Much of the data in the workers compensation claims process, such as the notes taken by claims adjusters, is unstructured. By digitizing this information, it can be easily searched using text mining. “We can look for words and phrases indicating a claimant’s apprehension with his or her progress toward the return-to-work objective,” says Moitoso. “Once there appears to be a problem, our people can step in with compassion and empathy to help solve the dilemma.”