Guest Speak Part 2: AI in Claims – Bridging the Operational Gap Part 1

Our world is fast-paced and changing rapidly. In the automobile sector, the number of vehicles on roads has increased including advanced vehicle technologies, and so has the unpredictability of weather and changing climate conditions. With an element of un-foreseeability thrown into everyday life despite certain predictable situations, insurers see an increase in payouts and claims complexities not just due to visible, tangible factors such as inflation and increased rates but due to intangibles like missed red flags during the first notice of loss (FNOL), and the lack of understanding of impactful facts of loss hidden in unstructured data.

Claims management has, for the better part of several decades, been a reactive process – repairing, resolving, mitigating, and replacing; rather than a proactive and prepared response to situations. InsureTech initiatives such as AI today, are helping bring into the mix a degree of predictability and pre-emptive response management in handling claims, managing cycle times, recognizing risks, etc. Technology platforms are building a data-driven risk mitigation environment for Insurers, with insights driven from past claim patterns to predict future trends leveraging institutional knowledge captured within AI. Many leaders have been focusing their efforts mainly around the recent paradigm shift from ‘repair and replace’ to ‘predict, prioritize and prevent’ when looking at AI-based Risk Management solutions.

Claims Director Brad Metzger, has worked in the P&C Industry for several years across diverse roles such as Claims Operations, Claims Strategy, and Claims IT. Having seen the rise of technology in claims management, he has a unique perspective in understanding and appreciating the need, to analyze and share claims data to understand emerging trends and risks. Below are his thoughts on how being proactive rather than reactive using tech-led data models, can greatly impact insurers and the decisions they make.

Reconfiguring the approach to claims management

There is no question that “predict and prevent” using customer-connected types of IoT devices adds value, but we see it as more of a redundant dependency or fail-safe kind of attribute. Many of the top 15 P&C Insurers in the US have implemented “predict and prevent” solutions. Still, as evidenced by soaring combined ratios, it is not yet proven to reduce overall claims frequency and severity results. As an aside, there is an argument to be made about the value in the Casualty space with “predict and prevent” vs. Physical / Material Damage, where most of the money and effort is going.

Many other insights need to be factored in when building out technology models to manage and mitigate risks. The most important of them in the P&C space is collaboration across different departments by sharing data insights. Access and insight to a wide range of data allows underwriters to learn from claims in real-time or near real time, and claims staff to make more informed decisions. Insurance advertisements today focus on why they should be chosen as the insurance carrier rather than educating analysts about various coverages, or impact of fraud and litigation, protecting against certain risks, weather event notifications and emerging risks. Insurance product marketing should be about establishing trust and connecting to their policyholders through claims knowledge, which is key to marketing. Actuaries would benefit immensely from deeper insights from claims, rather than traditional structured data based triangles for rate making and stat reserving.

Strategic Insights as the Differentiator

My most recent work using and sharing claims data to provide strategic insights was with Charlee.AI. is an up-and-coming Insurtech company using natural language processing, claims large language models and machine learning to derive deep insights using unstructured data and use them in predicting high severity, litigation, fraud, attorney involvement including efficiency in managing reserves. I led a pilot team employing Charlee on several use cases. Among the key learnings of using the InsureTech engine was that Charlee™ provided actionable insights through quantitative predictions on injury claims so that the representative could make informed decisions on predicting severity, likelihood of attorney involvement, and potential litigation. This lead to workflow changes, ensuring proactive management of claims, with high severity and expense risks and unusual patterns or behavior. The result of applying Charlee™ was compressed end-to-end cycle times, reduction in attorney involvement and, ultimately, severity and reserving accuracy. Reserving accuracy was not a KPI we agreed on measuring at the outset but ended up being a strategic insight and value-added consequence. Sharing the insights with the Enterprise Actuarial, Underwriting, and Pricing teams allows them to enhance several of their existing risk models.

Necessity of a risk mitigation environment

In my opinion, building an environment of risk mitigation for P&C Insurers involves a combination of advanced data analytics, technology, and strategic initiatives. Adoption is also equally critical. Getting key decision-makers to understand and prioritize effort and vision when it comes to establishing an environment of risk mitigation considerations tends to be a challenge. Especially as off late, the emphasis is more on near-term focus on results around underwriting profits and expense reduction to fuel growth. Below are other essential considerations necessary in creating an environment of data driven risk mitigation: 

  • Data collection and integration – Comprehensive and quality metrics from various sources (internal and external) and compliance with regulatory requirements are essential. This involves continuous updates of predictive models and risk assessments based on learning from across many carriers on current and emerging patterns.
  • Loss prevention risk engineering – Dedicated teams that employ data-driven loss collaboration with policyholders to prevent and mitigate loss cost. 
  • Predictive modeling – Making sure advanced analytic teams and leaders understand that models need to have a narrow focus on data that can predict claims frequency, severity, and emerging trends (e.g., social inflation drivers). This includes real-time analysis during the underwriting process to quickly and accurately assess both risk and price. 
  • Investment in technology – Identify and implement emerging technologies that support data analytics, AI, and automation to enhance mitigation and risk assessment capabilities. Many insurers do not fully embrace the need for tests and learning. And that it is ok to learn fast and fail forward.
  • Fraud Detection – Includes the ability to analyze underwriting and claims data for unusual patterns or behavior. The key here is early identification and investigation throughout the life of the policy. 


AI technology and its seismic impact on every aspect of the insurance ecosystem, especially in risk mitigation whether underwriting, pricing or claims management is obvious and far reaching. By ensuring greater data exchange and collaboration between departments promoting better transparency; leveraging data to enhance knowledge around possible and preventive claims scenarios; using technology as a ‘cyber supervisor’ to intervene at strategic decision-making junctures and boosting the capabilities of knowledge workers, AI adoption directly translates to claims risk mitigation. This in turn empowers institutional insurers to manage their operations from a ‘predict and prevent’ rather than a ‘react and restore’ standpoint.

In Part 2 of these series, we explore the topic of FNOL Standardization and how AI impacts the process.

Written by: Brad Metzger

Claims Director


Dr. Charmaine Kenita

Technology and Business Writer


What Products Are You Interested In?