In recent years, many insurers, especially property insurers, have dealt with combined ratios that have remained stubbornly above 100. Two primary factors are persistent inflation, which increases repair costs, and more frequent and severe catastrophic weather activity, which has driven higher-than-average losses for most of the past decade. This scenario has been exacerbated by reinsurers trying to maintain their bottom lines by minimizing their own risk exposure to severe weather or charging more for reinsurance. Without any ability to control inflation, and with little leverage over their risk transfer partners for the foreseeable future, what can insurers do to improve their combined ratios? Invariably, the answer lies in expense management, which is completely within an insurer’s control.
While approaches to expense management typically have involved outsourcing and reductions in force, technology has emerged as an equally favored approach in recent years. To be certain, insurers have been turning to technology to help with expense management for quite some time. For example, software that can perform a basic function such as bill generation, saving both time and money, emerged decades ago. However, in the past few years, technology that is much more sophisticated has emerged to assist insurers in tackling challenges that go well beyond bill generation.
Marketing
Expense reduction can start as early as prospect identification and applicant evaluation in marketing. In a perfect world, insurers could “pre-screen” prospects and only target those deemed worthy. Insurers that distribute through agents can rely on them as a pre-screening mechanism. However, insurers that are direct-to-consumer (or that use multiple channels and can sell directly), have more of a challenge because they often do not have clear visibility into applicant quality at this point. Exacerbating this lack of visibility is that these insurers tend to rely on demand generation tactics such as advertising, which can create and elevate overall brand awareness. The downside to this approach is that the messaging often reaches an unintended audience, leading to lower-quality applicants who have a lower chance of getting underwritten. When policies don’t get bound or get presented at a price that is too high for the applicant, that is a cost with no reward that insurers must bear. As this approach has become more expensive and has yielded mixed results, insurers have begun to reduce their advertising budgets and have begun to seek alternative ways to target a more desirable audience.
Many insurers have turned to targeted marketing to drive a more favorable applicant flow and reduce friction and costs in the underwriting process. These efforts often center on data mining, with insurer staff or third-party vendors culling through the data to develop a targeted list of desirable prospects. Generative AI (artificial intelligence) has emerged to complement these efforts by synthesizing the data and developing supplemental profiles of targeted prospects. Vendors such as Appian provide a generative AI tool that can be used to mine external data sources to profile existing and prospective customers, which will allow insurers to provide highly personalized content, products and services. These profiles can be used to guide messaging and product pitches based on information gleaned about the prospect, which could improve conversion ratios and lower overall acquisition costs. However, insights are only as good as the data they are based on, so forming profiles based on poor (or too little) data likely will fall short. In any case, robust profiles are no more than preliminary filters and are not substitutes for sound underwriting.
Another technology option for increasing efficiency and reducing expenses in marketing and lead development is an appetite alignment tool. By integrating with an agency management system, this tool allows clear communication between any insurer and its agencies about preferred types of business. Insurers enter their appetite preferences and changes into a user portal, and those entries will be communicated to their agencies. Without clear appetite communication, agents submitting out-of-appetite business that ultimately does not get underwritten wastes valuable resources. To be certain, leveraging both generative AI and appetite alignment tools can enhance the applicant pool, and the enhanced applicant pool in turn enable a more efficient underwriting process; however, they cannot solve all underwriting issues.
Underwriting
Underwriting’s primary goal is to make sound risk decisions that enable favorable loss ratios. However, insurers also can realize expense management goals in the underwriting process, especially when emerging technologies are leveraged to reduce lag time, increase straight-through processing and reduce costly underwriter referrals.
Generative AI can assist with all those tasks, but perhaps the most important role it can serve is that of a “completeness checker.” Given policy submission volumes, it is not easy for a human underwriter to verify that all the required information for a policy has been submitted in a timely and accurate manner. Generative AI can help solve this issue by flagging information gaps/inconsistencies, identifying missing components and validating submitted data against third-party data sources. This data validation can tighten up the underwriting cycle and reduce the amount of re-work and accelerate an underwriting decision. In May 2024, Hiscox USA announced a partnership with Convr AI to leverage its Risk 360 AI tool to ensure that underwriting and renewal decisions are based on the most accurate data possible.
Generative AI can also compare presented risks against existing risks within an insurer’s portfolio, help guide the overall risk determination for any application and make sure that the policy is priced appropriately. This is something that a human could certainly do, but the time it would take would be cost-prohibitive. At best, humans are able to make a few comparisons, which would not yield as complete a picture. In this sense, generative AI serves as a time saver but also improves loss ratios by flagging risks that are out of line with the overall portfolio.
Data itself can play a role in expense reduction, although any impact would be in collaboration with another technology (such as AI-driven predictive analytics). Efficiencies can be gained through data pre-fills, but the critical play with data is through analytics, which can assist with risk comparisons and largely affects loss ratios more than expense ratios.
See also: How to Predict Healthcare Costs
Service
Service has had the longest exposure to technology-driven expense management. Direct-to-consumer insurers always have had to provide service, and agent-focused insurers have taken on more service tasks from agents who chose to focus on revenue-generating activities. Policyholder service centers, viewed as cost centers, became targets for expense reduction through tactics such as outsourcing. However, outsourcing comes with inefficiencies and non-monetary costs. Insurers sought a scalable service option that could drive down transaction costs without sacrificing service quality, leading to the adoption of technology.
RPA (robotic process automation) has enabled rules-based chatbots to perform basic service tasks, freeing human capacity for more complex tasks. While RPA offers staffing relief, it is limited to low-complexity tasks, requiring human intervention for more complex issues. Generative AI, being more advanced, can interpret more complex questions and provide relevant responses, reducing the need for human intervention. Plenty of vendors, such as Boost AI, provide these types of services to insurers. Investing in service infrastructure cannot be avoided by direct-to-consumer insurers and may seem like an added burden to insurers that rely on agents. However, many insurers charge agents fees to offset service infrastructure investment costs. Both sets of insurers view service as a relationship enhancement that could lead to higher retention. Armed with technology, insurers can manage service costs much better.
See also: The True Cost of Big (Bad) Data
Claims
The final area in which technology can have an impact on expense management is claims. Some technologies are relatively pedestrian, such as claim “traffic management” systems that keep track of required steps and when they are completed, alert the person responsible for the next step that the step needs to be executed and send alerts when there is an information gap, to name a few. These prompts can save time and eliminate costly delays.
RPA-enabled chatbots can conduct an automated FNOL (first notice of loss) by guiding insureds through the process of collecting videos or photos of the damaged asset and can suggest re-takes if needed. In addition to conducting an FNOL, these chatbots can assign a claim to the appropriate adjuster, who can then begin the claim process. Having a claim assigned to the wrong adjuster is fairly common and can cause a great deal of delay and re-work, so avoiding this outcome will save money and time for any insurer.
While a chatbot-guided damage assessment is perfectly fine for smaller-scale damage (e.g., a single damaged asset such as an auto or a deck hit by a falling tree), it might not be optimal in the event of larger-scale damage resulting from severe weather events. For example, a chatbot might not be equipped to guide a damage assessment of a thoroughly destroyed asset. It also might not be possible for human adjusters to access the damaged property in a timely manner.
Another issue is staffing capacity. Many insurers rely on outsourced inspectors to supplement existing staff or stretch their existing staff to the limit and possibly lengthen the claim process. Both options come with costs (extra costs to hire the outsourced labor in the former; an increase in policyholder dissatisfaction and litigation likelihood in the latter). Clearly, insurers require a cost-effective inspection option that guarantees access to damaged assets.
To that end, many insurers have begun relying on aerial imagery providers as a scalable, less expensive alternative to live adjusters. Vendors such as Iceye can provide images that allow adjusters to assess severe weather damage without having to place staff onsite. These providers can cover a broad geographic region from the air, capture images of damaged properties in that region and send those images along to an insurer for it to synthesize. Insurers have a choice of image source (drone, plane or satellite), each of which has a varying degree of granularity. Without question, these aerial imagery providers can cover far more ground more quickly than a horde of live claim adjusters and can collect damage information at scale.
Once a human adjuster has these images, additional efficiency savings can be realized by leveraging AI. AI can compare collected images of damaged properties against pre-weather event images of those properties and can be trained to recognize the extent of asset damage much more quickly than a human can. This can be accomplished through a tool such as Swiss Re’s Rapid Damage Assessment solution. When integrated with claim estimation software such as Xactimate, AI can arrive at an initial damage estimate that, if needed, can be refined by humans. To be clear, these technologies are not being leveraged to replace employees in the claim process. Having high fixed-cost humans working on lower-end claims is expensive and consumes valuable capacity, and these technologies help insurers reduce expenses and increase staff capacity for handling more complex claims.
Another important technology available to insurers to help reduce expenses and increase staff capacity is fraud detection/prevention. This technology takes many approaches, but most often, the end product is a risk score rating the likelihood of a claim being fraudulent. This score may be based on a claimant’s history, recent transactions or known associates. The primary benefit to insurers is that if a risk score breaches an established acceptability threshold for fraud potential, they can apply extra scrutiny before the claim process gets too deep and is paid out. Without pre-emptive fraud detection, insurers must rely on their SIUs (special investigation units) to claw back money paid out for a fraudulent claim. This is known in the insurance industry as “pay and chase” and is not a great strategy for cost reduction or containment because getting money back is inherently more expensive than not paying it out to begin with. That said, despite an insurer’s best efforts, fraud is still possible, so it is best to have SIUs with as much capacity as possible to chase down payments that went out the door.
Not every technology discussed is going to be a fit for every insurer, but given existing expense pressures, it would behoove insurers to consider any available technology to help manage expenses.
