Automation is coming for commercial underwriters

As property and casualty insurers face margin pressures and increased competition, they are adopting automation capabilities to bolster customer-centricity, reduce costs, and improve operational efficiency, according to a May 2018 study by Capgemini.

From quoting to policy servicing, underwriting to claims processing, insurers are striving to improve the bottom line by automating processes across the value chain, the report titled P&C insurers add automation to their efficiency toolkits, states.

Automation is being powered by new channels for collecting customer data, advanced data processing capabilities, and improvements in artificial intelligence (AI) algorithms, the report explains. Because quick, error-free, and personalised service is expected these days as part of outstanding customer experience, automation is becoming a vital competitive edge and not just a cost-takeout play, the report adds.

In the commercial insurance space, robots are unlikely to replace underwriters any time soon, but the tasks and required capabilities are set to change rather quickly, says David Ovenden, global director, pricing product claims and underwriting at Willis Towers Watson.

“For a very large insurer that covers the whole commercial market, there will undoubtedly be portions of the portfolio that are entirely automatically underwritten but they will still have a guiding intelligence at the centre that manages that portfolio,” Ovenden says. “The need for this steering of how a product works will continue for the foreseeable future.

“As the data become scarcer and the risks become less homogenous, more complex and more different, underwriters will be needed but they will be supported by very much richer structured and unstructured data in an environment where the automated tools help them to make good decisions, very quickly.

“Underwriting will move to a phase where underwriters are pushed to make decisions much more quickly, but they will be better informed than they have ever been. The winners will be those who push the right data to the underwriter, at the right time to make the right calls in the right prioritised order,” he adds.

A winding road

It may be a long journey for some. Lloyd’s of London, for example, is working hard on the digitisation of the business, a prerequisite for automation.

The London Market Target Operating Model (TOM) is a core component of the market modernisation proposal, set out by the London Market Group (LMG), to make it easier to do business in the London Market, locally and globally. TOM is based on two principles: first is “one-touch data capture” in a common, global digital format, whereby data is entered only once, on behalf of all carriers and brokers, reducing risk of error and costly re-entry in multiple formats.

In order to accelerate the process, in March 2018 the Corporation of Lloyd’s introduced a mandate for electronic placement.

From the end of the second quarter of 2018, each syndicate is required to have written no less than 10 percent of its risks electronically. This target will rise by 10 percent each quarter until the fourth quarter of 2018, to reach 30 percent. Further targets will be confirmed prior to the end of the period.

“The London Market has not really moved on for a very long time,” Ovenden says. “TOM is making solid leaps into adding digital capability to an existing process but it hasn’t yet adopted the concept of non-human actors in the process, intelligent models that behave like underwriters or brokers,” he adds.

In order to take advantage of the opportunities that new technology offers, Lloyd’s has partnered with L Marks, an innovation specialist in the insurtech sector, and with the Boston Consulting Group (BCG) to set up and operate an innovation lab within the market.

The Lloyd’s Lab is set to officially launch in September 2018 and will focus on designing technology-driven solutions to meet the challenges the Lloyd’s market currently faces.

A new approach

Commercial insurance arguably requires a more tailored approach to clients than personal lines, but that should not prevent insurtech from offering a new approach to tackling risks in this segment, Trevor Maynard, Lloyd’s head of innovation, commercial, has told Intelligent Insurer in an interview. Instead, he believes that insurtech can help improve commercial insurance.

There are areas where technology can be both automated and tailored, Maynard explains.

“You can imagine a system that incorporates multiple data feeds from a client and uses them to offer automatically generated, tailored solutions,” he says.

He points to cyber insurance where insurtech firms use cutting-edge techniques, military intelligence techniques, and data feeds from many sources, and mash them together in clever ways in order to allow insights into all the risks and also the potential aggregations of risk across a portfolio.

There are, however, many barriers commercial insurers will have to overcome to get to this point, and in some areas it might never happen, Ovenden suggests. One challenge is the way commercial business is organised for large insurers, split into small, medium and large corporates.

“The number of employees is a metric often used as a basis to divide commercial insurance clients into small or medium-sized enterprises (SMEs) or larger corporates, but this is not necessarily a good measure of how sophisticated or risky the business is,” Ovenden says.

“The automation process does not depend on company size but on the variability and complexity inherent in the risk. Complexity is the main variable deciding whether a process can be automated or needs an underwriter to intervene,” he explains.

The way corporate client groups are divided into separate business segments means that for automation purposes, clients from different business units may need to be combined, Ovenden says.

While not all the risk in small and medium-sized enterprises (SME) space is suitable for automation, there is some in the larger companies that is, he notes.

“With automation you need to take out complexity and address repeatability,” he says.

Automation in large companies tends to be more complicated because of the large volumes of data, in addition to the higher complexity of the risk.

“The automation of large corporate risks is more about engineering the large quantity of case data. That might also be the case with the top end of mid-sized companies,” he notes.

One side-effect of the automation process may be that previously outsourced data processing may be in-sourced again.

“You don’t need lots of people re-keying things. That can be done by technology. This will effectively change the shape of the organisation,” Ovenden says.

Bringing repetitive processes back in-house also allows the insurer to reorganise processes to become more efficient and effective, he notes.

Two levels

There are two levels of automation, according to Ovenden. One is simple robotic process automation, when data needs to be moved from one system to the other.

“The act of integrating the data across a number of platforms means that you don’t have to re-key data; this in turn means you don’t have to employ people to re-key data,” he explains.

Then there is intelligent automation, when data is being moved from one system to another but the process includes a decision that needs to be made.

“Instead of having a one-dimensional rule saying ‘if the risk is bigger than X, send it to the survey system’, the insurer can look at the risk exposure system and run some large loss propensity models across that book and then start intervening in a more intelligent way, selecting the risks that have a tendency for large losses,” Ovenden suggests.

“A risk-scoring system can be used for operational intervention rather than just for pricing risk.”

As in the past when an underwriter and an underwriter assistant worked on a portfolio, Ovenden sees a similar setup in the future, supported by a digital underwriting assistant.

“Portfolio underwriters will have to be more analytically aware because people managing whole products will have a lot more data pushed at them. It’s changing the skillset needed on portfolio level,” he says.

The benefits of automation include lower cost, for example, as outsourcing is no longer needed, and also speed, with executives steering a portfolio more quickly using a larger amount of data. They will also be able to make better decisions, increasing the efficiency and effectiveness of operations, Ovenden explains.

“While the cost reduction driven by automation may become obvious quickly, the advantages of automation are more significant in the underwriting quality.

“Improvements on the expense base come more quickly, because if you have 500 fewer people on the payroll, the effect is immediate, while the loss ratio improvement has to earn through the portfolio, but actually the number is bigger,” he explains.

Insurers are starting to automate processes for small risks and are preparing to automate the pricing for these risks. However, there is still a lot of underwriter referral in commercial automation.

A few conditions are required to advance in this area, according to Ovenden. One of them is data integration: in order for this to be fulfilled the insurer will need to work closely with distribution to set expectations about automation and make it easy for the implementation to happen.

Furthermore, there has to be some broader thinking about data interoperability and expectations about when an underwriter needs to intervene and when this is not necessary. Another problem is that data assets in the commercial space are scarce, Ovenden adds.

“There is not the degree of accuracy as compared to personal lines. It’s significantly less because of the lower homogeneity,” he says.

Even with scarce data, insurers can make directional choices by setting intervention points, he notes.
There are other reasons for insurers to be hesitant about accelerating the automation of processes, such as the investment involved, Ovenden says.

“Everything is in a state of flux. It’s difficult to make a big bet on one thing,” he notes.