Insurance — Part 3: Claims Management
Insurance — Part 3: Claims Management
Machine learning and artificial intelligence can deepen retrospective analysis and ensure that decisions are informed by data, not subjectivity.
Join the DZone community and get the full member experience.Join For Free
Part 3: Claims Management
An insurance claim is "a formal request to an insurance company for coverage or compensation for a covered loss or policy event" (source: www.investopedia.com). Once initiated, the claim often goes through a complex process with one of two possible outcomes — the claim is either accepted, leading to a settlement, or rejected. The claims process would typically be: contact the insurance company, start of the claimant investigation, check the policy coverage, evaluate the damage and arrange compensation payment.
UK insurance industry figures are staggering. On average, £33m are paid per day in motor claims, £13m in property claims, £12.5m for policy protections, and £1m for travel claims; the average bodily injury claim is close to £10k; more than 98% of motor claims have been accepted, and the yearly cost of fraudulent claims is £1.3bn. Such massive claim expenses can lead to an underwriting loss, this is especially evident in motor insurance where an underwriting profit has only been made once in the last 24 years. (Source: www.abi.org.uk, 2017)
Clearly, insurers are faced with a number of challenges including high operational cost, constantly increasing customer demand, increased fraudulent claims and a lengthy process, hence customer dissatisfaction. Additionally, high IT costs, lag in change request, poor IT, and third-party integration increases the operational cost, which ultimately leads to an underwriting loss.
Despite the ongoing efforts of improving claims processes and fraud prevention, there is a scope for significant improvements focusing on better customer service and customer experience, improving operations and managing the claims more effectively, both in terms of time and resources.
To achieve these improvements we must integrate systems as best we can, and continuously incorporate the advances in predictive analytics, for example, and ambient computing such as GPS car tracking, telematics devices, body activity tracking and image recognition. Machine learning and artificial intelligence can deepen retrospective analysis and ensure that decisions are informed by data, not subjectivity.
One of the important areas of InsurTech innovations is support at first notification of loss (FNOL). FNOL, as the very first step in the insurance claims process, is often deemed a bottleneck of the process where a claims adjuster is faced with a number of challenging and sensitive issues, including the possibility a claim is fraudulent, policy coverage, and loss assessment. Failure to deal with the issues in a fair and effective way could damage customer relationships.
An FNOL decision support system can assist in this complex and time-pressure decision-making process using optimal resources. The FNOL eco-system is a real-time solution that typically utilizes a range of predictive models, artificial intelligence software, social and third-party networks, and ambient intelligence.
The utilization of predictive models typical for motor insurance is illustrated in Figure 1. Models are deployed on a scoring engine as RESTful web services and connected to a front-end. They are simultaneously scored in real-time during the first notification of loss and model scores are visualized on the dashboard (Figure 2).
Figure 1. FNOL Decision Support Tool in Real-time
The choice of predictive models depends on policy type and insurers' preferences. Motor insurance models typically include predictions of a vehicle being written-off or recovered; A bodily injury severity score would assist in more accurate predictions of the estimated claims reserve amount; total settlement cost would give total estimated claim cost, and a fraud model would flag any potentially fraudulent claim. Additionally, the fraud model could be designed using an existing company's fraud methodology and implemented as a composite solution consisting of a series of models and fraud indicators.
Figure 2. FNOL Customer Support Dashboard (prototype)
Represented on an intuitive dashboard the FNOL solution acts as a handrail, assisting claims adjusters in better resource allocation and task prioritization. Decisions based on the dashboard indicators such as an early settlement cost offer can significantly cut unnecessary administrative cost that would otherwise have been incurred.
Dynamic questionnaires can be implemented to direct the claims process by creating a bespoke strategy based on the dashboard scoreboard (that is, model outputs). For example, if the fraud gauge is "flashing red" then a different set of questions can be pulled out from the database to probe the claimant. On the other side, a "steady green" light would skip additional questions making the claim process faster and ultimately increasing customer satisfaction.
With the overall tendency towards more customer-centric services, the cutting-edge InsurTech innovations are focused on self-service solutions available anywhere, at any time. The latest FNOL tools utilize automated insurance agents (known as claimbots) that make conversation, exchange information, and make assessments and recommendations often faster than humans do.
Even though it has been around for several years, the potential of the FNOL decision support tool has not been fully exploited. A fraction of insurers utilize the tool in its simplest form (Figure 2). Market research shows that insurers are willing to adopt AI technologies, however, cost of implementation and integration, data security, and imposed regulatory rules have been the main stoppers in these endeavors.
Given the necessity of making the claims management more effective, the urge for such support tools becomes more evident. The promising news is that baby steps can be adopted that would gradually lead to a robust and superior claims management decision support ecosystem.
Published at DZone with permission of Natasha Mashanovich , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.