Selecting and pricing the right risk no longer a carrier's conundrum

How do you select the right risks and price them appropriately? That’s the goal of insurance underwriting, but as we all know, there’s a delicate balance between charging acceptable premiums and achieving target profitability.

While the science behind it is valid, little has changed in the way underwriting is performed by commercial insurance underwriters. This process is really hard because no single source is reliable, transparent, or comprehensive for assessing current or future risk.?

Understaffing exacerbates the situation. There are dozens of cumbersome tasks and weeding through voluminous data sets and sources. Old underwriting tools require manual key-in submission information. Required quality checks contribute to even longer lead times and added stress to maintain a high underwriting standard.

The demand for faster submission processing and improved customer experience always has carriers looking for ways to speed submission to quote. Insurance providers want straight-through processing, while customers demand real-time answers, quotes and binders.

New AI technology has modernized the process of risk scoring. Structured and unstructured data from thousands of data sources and billions of data points are used to verify, cross-reference and deliver critical underwriting insights in real-time.

Today, with machine learning models AI can produce a single number measure of risk on a scale of zero to 100 for businesses that an insurer can look at, quickly assess the relative risk of insuring that business and prioritize submissions. Companies that return a higher score would be riskier to insure while those with lower scores would likely present less risk: an underwriter can quickly see that a score of 90 means only 10% of businesses are riskier, while a score of 30 means 30% of businesses are less risky. The higher the score, the higher the risk. This process is as simple as filling in the name and address of a business and seeing real results that can affirm decision-making.

An example from commercial auto

Commercial auto is a line of business where risk scores can be especially valuable for insurers. Underwriting profit in?commercial auto ?had been trending down for several years. In a study my company conducted as depicted below, AI modeling indicates commercial auto loss experience at the industry level were functions of (mostly) many geo-temporal macroeconomic and firmographic variables. Additional models looked at commercial auto litigation verdict amounts and large individual losses. Then models were trained to forecast injuries, accidents, fatalities, violations and more, per power unit.

Ultimately, modeling accidents, injuries, and fatalities at the business level in a forward-looking way resulted in the most valuable predictive model(s) for underwriters:

A recent retrospective study looked at our proprietary risk models as of March 31, 2022, for all U.S. businesses that had a Department of Transportation number. We then recorded how many accidents, injuries and fatalities occurred for these businesses in future months (April – October), and then aggregated occurrences per power unit (shown in vertical axes in figures below) by risk score (shown in horizontal axes in figures below):

The results showed a convincing (and expected) relationship between risk score and future accidents, injuries and fatalities per power unit. Businesses with higher scores experienced more of these events, on average, and businesses with lower scores experienced fewer of these events, on average. The study shows clearly that insurers with access to AI modeling would have been able to make decisions about current and potential insureds that would have resulted in better loss experiences in the year ahead.

Within the?commercial auto ?example, customers who were writing these policies have been avoiding the riskiest insureds or those with the highest scores. They have been able to prioritize the best risks (businesses with the lowest scores). And they have been able to charge more (less) premium for insureds with higher (lower) scores.

For carriers,?AI-powered risk modeling ?can enable better risk decisions, streamlined risk decision-making, and optimized underwriting–with faster quote-to-market. If thinking about implementing AI modeling to better inform underwriting, here are five key advantages to consider:

1. Increased underwriting productivity and speed to quote

Prioritizing and reviewing submissions is often very manual and cumbersome. Instead, underwriters can rapidly narrow risks within their appetite and deep dive on selected risks.

A systematic approach reduces the administrative burden on underwriters and reduces the time to quote. Crum & Forster Insurance cut its surplus and specialty (S&S) lines operations submission processing time by half, enabling it to redeploy 40% of its contractors to higher-value roles and 20% of operations employees to revenue-generating roles.

2. Reduce underwriting operating costs

Automating the operational steps required for clearance and underwriting file preparation can significantly lower operational costs (FTE or BPO). For example, Columbia Insurance shaved off 20 minutes on average per submission, or a total of more than 2,500 hours per year in reviewing submissions.

3. Calibrate risk selection

As mentioned above, a single number measure of risk scoring, on a scale of zero to 100, can be used to evaluate the relative risk of insuring that business.

4. Pricing adequacy

By monitoring loss performance risk scores, customers can identify segments of their book of business where they have an opportunity to adjust pricing to better reach their target loss ratio

5. See growth of premiums

Increased underwriting productivity and speed-to-quote drives increased quote ratios, resulting in increased binds and new business. For example, Columbia Insurance Group reduced straight-through-processing on SMB policies by 28% within three months.

To learn more about Convr visit convr.com .