As per the report by a leading technology consulting firm, 75% of insurance executives believe Artificial Intelligence (AI) will transform or bring significant change to the industry over the next 3 years.
A leading life insurer in Japan believes that AI will potentially increase its productivity in claim payout area by almost 30%!
The above statistics provide a sneak peek into the insurance world and how AI will make its way into the insurance industry. With this blog, let us understand how insurance claims management is re-defined and re-lived from the perspective of digital disruption, in general and artificial intelligence, in particular.
First Notice of Loss (FNOL) - Reporting a claim:
Depending on the maturity of AI technology, a claim can be reported, routed, triaged and assigned, with or without any human intervention. Virtual assistant also known as Chatbot laden, along with Natural Language Processing (NLP) and/or speech recognition technology can efficiently handle First Notice of Loss (FNOL) reporting process thereby, increasing productivity and making processes lean.
This, coupled with Machine Learning and Deep Learning will lead to bots assigning claims to adjusters basis on their workload and expertise. This in-turn will create value in the form of low turnaround time and reduced claims leakage and bring in efficiency into overall claim portfolio management.
Given the pace at which AI is growing, the day is not too far when IoT-enabled sensors fitted in your car or home sends distress signals to your insurer informing them of the accident, prompting a bot to open a claim. It could possibly even connect with drones surveilling that particular area, placing a request to feed live pictures of the accident scene and passing the information on to an adjuster. And, some of the leading global insurers, IT vendors and insurtechs are already focusing on creating POVs as well as pilot projects in these areas.
Claims Investigation made better through AI:
The traditional approach of investigation to identify suspicious claims required a lot of manual effort and time to analyze large amounts of data (online/paper documents) and also closely monitoring and/or following the claimant for suspicious activities. This approach may fail due to lack of time available to get the job done or to investigate. Insurance companies cannot afford to hire more investigators to speed up the process as that would add up to their cost.
In such cases, AI based investigation tools play a key role by using the external data sources from social media platforms, investigating bureaus, police reports, ISO claim history reports and so on. The data is extracted to build a timeline of events to support or dishonor the claim. The AI component cross references the social media profiles and aggregates the data to form any correlation between the claim and the people associated to identify any suspicious or fraudulent activities.
Enhanced Loss estimation for reduced claim leakage:
Less than 10 years ago, we humans walked to the bank to deposit a cheque in a bank account. Now, it is more common to upload the photo of the cheque using a mobile app to deposit it into the account. Can this be done in an Insurance claim scenario too? Yes, it is happening now through AI technologies. Deep Learning (DL) algorithms with image recognition model helps predict or assess the damage, based on the photo of the damaged vehicle. The model can predict whether the vehicle should be towed to the nearest repair facility or to the salvage yard.
For example, when a repairer orders for a front bumper with an energy absorber and a license plate frame, the system suggests that there is a great chance of front bumper cover clip that may also be needed to complete the repair. So, this way, AI technologies help to better assess the potential loss and suggest the parts that are likely to be repaired, to estimate the loss quickly and efficiently.
Impact of AI on Claim Reserving:
The aggregated claims (traditional) approach for reserving often neglects the detailed information of an individual claim like claims reported date, claim diagnosis, lawyers involved and so on. So, there is always a chance of missing a fraudulent case by following this aggregated approach. The Individual claims based along with AI reserving model analyzes the information and happenings around each individual claim and associated claimants to suggest any change in reserves automatically before the settlement is initiated.
Fraud Detection through AI:
Fraud detection is all about connecting the right data points to discover any suspicious activity before it emerges as a big threat. Understanding the interactions between products, devices and locations, the mapping of these associated data points to users, clients and employees helps identify the fraudulent behavior. With the innovation of technologies, there may be many fraud detection techniques in use. But, the technique to be used solely depends on the maturity of the organization and the level of fraud data it deals with.
Fraud detection techniques in use in today’s insurance market:
Fraud detection techniques vary depending on the organization and the fraudulence level. Listed below are a few techniques that are prevalent now.
Unsupervised and Supervised Machine Learning:
The technique in which the software or the model learns from the large amount of data without giving any information about the data is known as unsupervised machine learning. This technique helps the AI model to self-improvise by getting exposed to diversified set of data. The algorithm analyzes and segments the data based on the relation or patterns that exist among the data. It helps to segment the claims data based on the geographical location, third parties such as lawyers, vendors involved, any suspicious terms identified from adjuster’s notes and so on. After segmentation, a human intelligence is required to assess these claims for fraudulence.
These categorized claims are then fed to supervised machine learning algorithm and it is expected to learn these patterns and predict for any similar patterns in the future. This approach helps to predict the fraudulent activity based on the existing patterns and in turn identify new fraud patterns in the long run.
Harnessing Unstructured Data:
There are a large amount of unstructured data collected during the entire claim process through different sources such as adjuster notes, e-mails, information collected through phone calls, interview with claimants, third Party reports and so on. These data may have hidden suspicious information useful to identify a fraudulent claim, which may not be caught by humans during investigation. But, unfortunately these are not directly stored anywhere in data warehouse and just available as paper documents, notes, emails or reports. In this scenario, text analytics play a leading role in analyzing these data through text mining software to derive meaningful insights to get a better understanding of the claim. This software looks for any suspicious or scripted events or terms in the claims to identify any discrepancies or fraudulent conditions.
To conclude, the benefits of AI for claims management are many, it can optimize claim lifecycle, reduce claims cost by automation and lean processes, enhance client satisfaction and in-turn keep the insurance prices affordable. But, it is still at its nascent stage and has a long way to go. It will be exciting to see if traditional insurers and industry disrupters can play their cards and turn the tide in their favor or create a space for both the parties to take the fair share.
Do you agree with us that AI is going to benefit the insurance industry in the future?
Know more about Mindtree's Insurance services and solutions
To know more about Intelligent Automation for the Insurance Industry and how automation is changing the insurance industry landscape do click the link