How Automation Changes the Role of the Underwriter for Low-Complexity Risks

As the insurance industry evolves, underwriting organizations are under pressure to make more informed decisions and stay competitive in the industry. One way to do this is by investing in data and using it to automate the underwriting process for low-complexity insurance products.

In this article, we’ll discuss how data investments can enable automation in underwriting, transforming low-complexity products into “low- or no-touch” products.

Benefits of Automating Low-Complexity Underwriting

Low-complexity underwriting refers to insurance products that have a relatively straightforward risk assessment process, with a limited number of variables to consider. Examples of low-complexity insurance products include personal auto insurance, term life insurance, and home insurance. These products typically have a well-defined set of underwriting guidelines that can be automated with pre-defined business rules for elements such as appetite and pricing.

By automating these low-complexity products, underwriting organizations can free up underwriting professionals to spend more time evaluating edge-case opportunities, rule exceptions, or more complex lines altogether. In other words, the underwriter’s role shifts from being the primary decision maker to being more of a “reviewer” of deals. This allows underwriters to focus on higher-value tasks, such as evaluating risks that fall outside of the pre-defined business rules and identifying patterns or trends that may require adjustments to the underwriting guidelines.

This new role of “reviewer” of deals is different from today’s underwriter, who is often the primary decision maker on all deals, with the help of some guidelines. By automating low-complexity products, underwriters can focus on complex cases, and bring more value to the organization by reviewing the automated underwriting decisions and identifying areas for improvement. This allows them to focus on more strategic tasks, such as identifying new business opportunities supporting the development of new products. 

Automating low-complexity underwriting can also deliver cost savings, as it reduces the need for manual data entry and eliminates errors that can occur during the underwriting process. This automation inevitably leads both to faster underwriting decisions and to more satisfied customers as the process is streamlined and automated.

The shift in focus from tedious day-to-day underwriting to reviewing automated decision-making and taking on higher-value tasks can lead to more efficient and accurate underwriting – and, ultimately, to a more profitable and successful insurance organization.

Compliance with Regulations

Before diving into the human challenges in this space, it’s worth explicitly defining what we mean when we talk about automation. To automate underwriting, you can impose a set of common rules for appetite, rating, and pricing, and you can trigger approvals and notifications on items that fall into borderline acceptance for attention. That has been done for simpler lines for a while, and doesn’t require AI to put into place. AI comes into play as you look at the outcomes of the policies written over time and introduce new or dynamic rules to reclassify appetites, ratings, or pricing models. This introduces new challenges, as the judgement criteria evolve over time.

One big risk with this sort of AI-driven automation is that regulators are not yet in a place where they are comfortable with insurers using machine learning to dynamically improve their ratings and experience mods. To avoid regulatory non-compliance, organizations should focus on following appropriate regulatory filing guidelines and on improving and automating manual processes with clear sets of rules that are followed uniformly. If insurers begin to bleed into pricing even simple risks with machine learning – to the point where they can’t clearly articulate how the decision was determined – they are at severe risk of regulatory non-compliance. You can use analytics and data science to change your rules, but those rules (at least for now) are likely best separately managed and clearly explainable.

Challenges of Automating Low-Complexity Underwriting

Even if you handle the AI aspect clearly, automating low-complexity underwriting comes with its own set of personnel challenges – one of the biggest being the cultural change needed to better manage underwriting professionals’ perceptions on their role and the trustworthiness of the automated underwriting decisions. Without intentional, holistic change management, these AI implementations are destined to fail.

This change management largely should consist of a combination of proactive communication, collaborative design, the re-imagining of the definition of the role, and strong consideration of the structure of incentives. If insurers intend to disrupt the way underwriters work, they must ensure they do not do so lightly. Proactive communication is essential in ensuring that underwriting professionals understand the benefits and limitations of the system and how it will impact their role.

Likewise, collaborative, participatory design reinforces this idea by ensuring that underwriting professionals are involved in the design process, and that their input is taken into consideration when developing the system. By bringing underwriters along for the ride and creating internal advocates, insurance organizations can seed adoption and buy-in through a network effect, which can be a powerful force. Additionally, re-imagining the definition of the role of the underwriter in concert with these investments lets the business consider how the role of the underwriting professional can evolve and change with the introduction of the automation system.

Perhaps most importantly, it is important to recognize that individuals tend to perform to what they are measured on. Underwriting professionals may lose agency when they defer decisions to the computer, which means they are often in less control over their incentives. To unlock the potential of automation, there is a need to adjust what incentives are reviewed and how incentives are managed. For example, an underwriter may currently be incentivized on total premiums booked, which may lead to them preferring to write less desirable risks to achieve their quota goals. Meanwhile, an AI system may prioritize portfolio profitability optimization (perhaps utilizing predictive analytics projecting loss forecasts) and suggest walking away from a number of deals. In this scenario, the UW is actively incentivized to comb through the tool’s recommendations and look for potential “exceptions” that they can justify as mis-characterizations so they can write more work and achieve their goals. This example illustrates how, if not properly considered, the structure of incentives may create a barrier to adoption and acceptance of the system among underwriting professionals.

Overcoming the Trust Hurdle

Even if underwriting professionals do buy into the conceptual changes with proactive change management, there will still, at first, be some doubt and skepticism about the quality of the automation. To overcome this, it’s important to provide clear data informing a reviewer of key decision criteria, time/date stamps of data pulls and decisions, and escalation paths. Additionally, it is critical to have laid a strong data foundation – not just by gathering the needed data, but by ensuring the management strategy and infrastructure is appropriately tooled to support the system with accurate, timely data. Finally, it’s also important to take advantage of the expert underwriting professionals who will be auditing the tool and let the system and processes learn from their feedback. Trust is hard to win, but with the appropriate transparency (and time) eventually users will verify enough decisions to begin to leverage the recommendations more confidently.

The Path Forward

Data investments can enable amazing automation in underwriting for low-complexity insurance products. In doing so, organizations can improve decision-making, increase efficiency, and reduce costs. However, it’s important to be aware that taking on this automation effort also comes with its own set of challenges:

• A cultural change is needed for better management of underwriting professionals’ perceptions on their evolved role in the new business system.
• Time and patience are also needed (along with good technology foundations) to drive the trustworthiness of the automated underwriting decisions.

To overcome these challenges, organizations should focus on providing clear data, adjusting incentives, and taking advantage of feedback from expert underwriting professionals. Additionally, organizations should be aware of the regulatory risks and focus on improving and automating manual processes with clear sets of rules, rather than setting an AI model free to run wild on the underwriting process.

By addressing these challenges, organizations can successfully transform low-complexity insurance products into “low- or no-touch” products, improving their efficiency, decision-making capabilities, and overall profitability.