How Machine Learning Is Transforming Insurance Claims | PropertyCasualty360

Data is the lifeblood of any machine learning model. (ipopba/Adobe Stock)

Machine learning (“ML”) is one of the most prolific areas when it comes to impactful use cases for the insurance industry. Also, within insurance, claims management offers one of the most promising areas to apply this technology due to the large amount of data available for training algorithms and the consistency of the principles applied in the claims evaluation process. . Here we look at some use cases for ML in the claims and challenges that limit adoption.

focus on fraud

fraud detection Perhaps the fastest growing area of ​​ML adoption among insurers.startups such as shift technology, Frith When Owl Institute Seeing strong demand from carriers, it has raised significant sums of money from investors to support its growth. These tools work by applying cutting-edge data science to large historical billing data sets enriched with third-party data.

Insurance companies have been quick to adopt ML-based fraud detection strategies. This is because it often provides the most immediate and tangible return on investment. Many claims fraud teams rely on a rules-based approach (often in the head of an assessor) and tend to focus on the most obvious claims where the scope of fraud is larger and more complex. fraud cases may be ignored. In contrast, ML can often detect more subtle patterns of fraud that may not be visible on individual claims. For example, does one motor body shop routinely overcharge multiple claims? Does a doctor’s office routinely diagnose whiplash patients in low-impact accidents? Patterns may only become apparent when reviewing all past claims related to individual policyholders or vendors.

Assessor automation

Additional areas of significant innovation activity include: damage assessmentThe development of computer vision for analyzing billing photos has become a core innovation for several start-ups, especially in the personal space.

In a motor claim, companies easy to handle When snap sheet Enables policyholders to submit photos of damaged vehicles and human assessors to create claim estimates without looking at the images. Their models have been trained on similar historical vehicle data, often enough to resolve claims quickly or make informed decisions about how best to proceed with claims. Accurate enough. Computer vision evaluations speed up decision making, reduce billing leaks (making large payments to settle claims), and improve the customer experience. Expert evaluators still have a role, but they are directed at less reliable and more complex cases.

Similar products are appearing in property damage assessments from startups such as: fly reel (acquired by Lexis Nexus), hoverWhen Hosta Institute.

document processing

A new area that seems to have enormous potential is using ML to automate the processing of complex documents. Claims professionals today are buried in documents. Even simple claims can feature damage reports, doctor’s notes, multiple bills, emails, and texts. They all contain important information in unstructured or semi-structured form.

Natural Language Processing (NLP) and computer vision technologies have the potential to significantly reduce manual data entry. For example, an ML application can look at an invoice and extract individual items, payment information, and invoice number. At the end of the process, the accounts payable clerk just needs to approve the invoice with all the necessary context at their fingertips. Small claims that meet certain criteria and have no “red flags” can be approved without human intervention.

Startups often train models on a single class of documents (at least initially). Hypatos Invoice processing leader. digital owl Emphasis is placed on reviewing medical records. ground speed analysis Started as a Rothrun specialist. Insurers can combine these tools to extract all the information they need from documents, allowing claims professionals to focus on decision making rather than data entry.

Barriers to adoption

In other words, new companies are emerging that are using ML to streamline billing management. Perhaps the single biggest challenge these companies currently face in adoption by existing insurers is integration into existing systems.

Data is the lifeblood of any machine learning model. ML applications require historical data to train and tune models. And once you’re up and running, you need to quickly access and activate new billing data. Currently, only a few carriers have the technology to deploy ML models for insurance claims operations.

There are many potential use cases for ML that can streamline the claims management process. However, many start-ups have stalled on integration with carriers involved in the first pilots, and projects never move past his POC stage. To unlock the full potential of ML in claims, insurers must transform their core IT systems.

Based in London, Jack Prescott is a senior associate at MTech Capital, a venture capital fund focused on the insurtech space. Jack Prescott

Based in London, Jack Prescott is a senior associate of mtech capital, A venture capital fund specializing in the insurtech field.

See also

Leave a Comment