Listen to this story
My last article was about how insurers can leverage AI & analytics in underwriting. This article focuses on the next step of the insurance value chain, which is claims. We all have experience in claims, but we hardly know how it is processed at the backend and how AI is bringing more efficiency and effectiveness.
Claims is a formal request made to your insurance provider for reimbursement against losses covered under your insurance policy.
Sign up for your weekly dose of what's up in emerging technology.
Digitising insurance claims
Today more than half of claims activities have been replaced by automation, according to McKinsey. For example, insurance providers have been using advanced algorithms to handle initial claims routing, increasing efficiency and accuracy.
IoT sensors and an array of data-capturing technologies have helped insurers replace manual methods of first notice of loss (FNOL), where claims triage and repair services are often triggered automatically upon loss. For example, in the case of an auto accident, the policyholder takes a streaming video of the damage, which is later translated into loss descriptions, and estimated amounts. At home, IoT devices are used to proactively monitor water levels, temperature, and other risk factors and will proactively alert both insurers and tenants of issues before they arise.
Meanwhile, automated customer service apps handle most policyholder interactions through text and voice directly following self-learning scripts that interface with the fraud, claims, medical service, policy, and repair systems due to faster resolution times. According to Accenture, nearly 74 percent of customers said they would interact with modern technology and appreciate the computer-generated system of insurance advice.
Human claims management currently focuses on a few areas. This includes complex and unusual claims and contested claims where data-driven insights and analytics empower human interactions.
In this article, we will discuss how insurance claims processing is getting transformed by the adoption of AI, NLP and analytics. As AI and NLP enable digitisation, the power of analytics boosts the effectiveness of extracted content. This is done by leveraging predictive modeling.
First Notice of Loss (FNOL)
In claims management, the first process is FNOL, where the insured informs their insurer about the loss and lodges a claim with the insurer for the damages incurred.
Lodging a claim can go through multiple channels. This includes voice/on-call, non-voice-through portal, mobile application or sending an email. In most non-voice cases, the lodgement process is keying in 60-80 fields (insured details, vehicle details, loss details, customer details, etc.) in the claims management system (Guidewire, LexisNexis, etc.).
The information is read from emails, standard and non-standard claim forms, and other supporting documents that are key to the claims management system. These forms can be digital, handwritten, scanned or unscanned, and can be of various formats, including .eml, .msg, pdf, Docx., .rtf, TIFF, JPEG, etc.
|Industry specific||Insurer claim forms|
|Insured specific forms||Created by larger brokers|
|Supporting documents||Police reports, incident reports, legal forms|
Here’s how it’s done
The key steps involve automating the 60-80 features with AI. The nature of the fields is not standard; there would be free-text driven fields (claims description, type, etc.), which require ML and NLP models to classify and summarise to decide the loss type. Or, there would be documents such as police reports and survey reports which vary based on city and state and are mostly handwritten.
To yield significant impact, one needs an advanced pre-built ensemble of computer vision and NLP models that can extract, classify, and summarise details.
Further, the extracted information enables the building of ML models to identify duplicate and fraud claims. For example, the claims can come from the same customer again and again, or the same agent is sending across similar claims. All such cases are identified and sorted through ML models.
Lastly, the AI analytics’ intervention would be to direct the claim to the correct channel for processing and prioritisation. Alignment can be either auto-processing or handler driven depending upon the claim complexity, value and ageing. The models support in directing the case to the right-skilled handler.
Once the automated data pipeline has been established, you can leverage the rich, newly aggregated data to get powerful insights, leading to better underwriting decisions, products and customer experience.
Benefits of digitising claims set-up processing:
- Efficiency gains 30-50% due to manual work reduction
- Better customer experience with faster claims set-up
- Identification of duplicate claims
- Faster decisioning with automated claims adjudication
- Revenue prevention by proactive fraud detection
- Further insights: Claims volume prediction, loss analysis
- Reduce chances of litigation by faster and more accurate processing
Setting up claims and proceeding with the same can be a long, effort-intensive process which can lead to revenue leakage. Applying AI and analytics in claims management improves cost optimisation and customer experience and curbs revenue leakage. Companies should adopt AI solutions which are domain-rich, scalable and flexible in delivering multiple use cases.
This article is written by a member of the AIM Leaders Council. AIM Leaders Council is an invitation-only forum of senior executives in the Data Science and Analytics industry. To check if you are eligible for a membership, please fill out the form here.