Artificial Intelligence (AI)… The phrase and concept that transitioned from predictive modelling (inclusive of loss rating / experience models) of yesteryear to The Terminatorof tomorrow. While the debate continues as to the value, predictability and credibility of AI, the fact is that tomorrow is here today. No doubt that AI will continue to evolve into a better product gaining greater consumer confidence.
That said, many insurance products can use AI with some level confidence. I would categorize these products as first dollar (fully insured) and claims frequency driven, such as life, personal lines and commercial coverages that are not subject to risk retention, such as high deductibles, self-insured, excess, etc. Generally, businesses that elect a risk retention avenue to their insurance needs are confident risk takers understanding the financial benefits and challenges of such approaches.
One of the challenges with AI is either how it’s packaged and/or perceived application by the user. In recent years, the tendency appears to have AI replace the underwriter in part, if not in whole. I can appreciate the “why” due to the lack of available qualified talent in the industry and/or limited interest in entering into the insurance industry. To that end, the application of AI might be acceptable on small account portfolios, high volume and high claims frequency insurance lines, however it’s more difficult to utilize such technology on claims severity driven and risk retention products.
It must be acknowledged that skills, methodology and performance assessment will be dependent upon at what level of the organization is being discussed. From the top down, it’s macro to micro. That is, at the home office at the chief underwriting officer’s level, the analysis should largely be on the product line’s performance with an understanding of impact to the portfolio due to economic, market conditions, geographics, and other high-level variables. At the other end of the spectrum, especially for middle market and larger accounts, it’s based on the individual risk loss experience and characteristics, market pressures, production relationships and other considerations impacting the underwriter on the battleground. This article will focus on the underwriting on the frontlines.
While the techies may disagree, I do, however, believe that AI will have its limitations – namely it lacks the “gut instinct” or “six sense.” I don’t think AI has evolved enough to that level and respectfully, may never in certain aspects of insurance. For the insurance industry, this distinguishes the human vs AI underwriting and claims handling. These intuitions come from experiencing outcomes that challenge the analysis of relying on solely quantitative data. In my opinion, experience and skill are the apex instrument in effectively evaluating, blending and balancing quantitative and qualitative information to produce opportunities or avoid situations that would have otherwise been missed.
Stepping back several decades – on the front lines of underwriting (individual account underwriting), loss rating or loss experience models (model) were widely used and started to transition from high frequency predictable product lines to severity prone products with less credibility. Nonetheless, the success of these models on predictable and credible products relied on the experience and skills of individuals who were able to read the tea leaves of the future, thus tweaking the models accordingly for the benefit of the end-user (field underwriters), who used the model only as a tool not their ultimate decision maker. Underwriter discretions and deviations from the models were (and should be) common practices. In my experience, those who used the model as the end all, often missed identifying good and bad risks, therefore increased the chances of having an adverse underwriting portfolio.
Underwriting leaders often conclude that a poor performing risk or portfolio was due to an underwriter’s practices, when they should be evaluating whether the risk was poorly underwritten as opposed to underpriced. Underwriting leaders also fail when they determine an underwriter’s success via adherence to guidelines or authorities by measuring variables such as discretionary pricing exercised, minimum and maximum premium size, minimum three-to-five-year acceptable loss ratio and other similar metrics. To that end, while leadership can impose various underwriting limitations via guidelines or authorities, of importance is the need to train then trust their underwriters to make tough risk selections decisions while always aiming for rate adequacy.
Where does the art of underwriting come in? While accounts can be underwritten via a model, the challenge is when there’s an outliner or anomaly to the risk, which can be quantitative, qualitative or both. That said, to the extent possible, a good underwriter would be inquiring about such information in advance to either discount, surcharge or avoid the risk. For example, at the simplest form, an underwriter will be able to assess an outliner policy year, which may drive the anticipated rate need in the opposite direction than the actual need. An experienced and skilled underwriter will be able to assess if the policy year in question is the result of claims frequency and/or severity and possible underlying cause(s) to the activity. This can also be said for mid-term and pre-renewal underwriting, which generally is employed with middle market sized accounts and larger via the following triggers – paid / pending losses, claims, premium audit, risk management (aka loss control), account management, producer (aka agent / broker), etc. Underwriters’ value is further elevated for those that have the ability to develop relationships, negotiate and sell, which is often overlooked as required attributes.
The value of a seasoned underwriter can be immeasurable with risk retention products. These underwriters must be able to fully capitalize on quantitative and qualitative equations for the best underwriting outcome largely because the products they engage involve the policyholder retaining risk, while transferring potential large losses to the insurance carrier and/or reinsurer. While some risk retention insurance products may have greater predictability than others, ultimately due to the claim severity expectation, these claims would be fortuitous in nature, therefore difficult to predict. These underwriting skills increase in value during soft market cycles due to downward pricing pressures and increase premium production while often working with less-than-ideal underwriting information. Generally, these underwriters place a higher level of proactive portfolio management – as if they were managing multiple investments for an aggregate profitable return.
Underwriters engaged in self-insured, excess or reinsurance product lines must be skilled, and those who are seasoned and demonstrated the capabilities must be entrusted by leadership to produce the best outcomes with minimal oversight. For underwriters that fit this description, I’ve found three key measurements to establish their success – renewal premium retention, new business production and profitable gross loss ratio. Management may need to temper these measurements subject to current market conditions to remain viable and competitive.
Now back to the title of this article… While models and AI are welcomed, they should be used only as tools, especially on complex and risk retention products require a unique and special set of skills. Overlooking the need or worse overriding the skills with models and/or AI, will most likely produce an adverse portfolio. We should not undervalue a seasoned underwriter’s instinct to identify profitable opportunities regardless of the model’s output. This is equally, if not more important, to address if not avoid the “perfect” quantitative account, because there is qualitative information that suggests the risk is heading south. I believe these skills are learned and instinctive, thus requires the feel similar to the concept of Kentucky windage.
Joseph Dore is president of USBenefits Insurance Services, LLC.
