Getting Real with AI in Insurance
The last several years have shown us that we are in the midst of an artificial intelligence (AI) revolution that will likely transform almost every aspect of our lives over the coming months, years and decades. In few areas has this become more apparent than in the property & casualty (P&C) insurance industry. Today, start-ups and legacy providers alike are turning to AI to gain actionable insights into their current customer base, expand that base, and gain back office efficiencies in the areas of underwriting and claims management. As AI becomes more ubiquitous across the industry, the need for tight governance rises.Today, we will cover these areas in what we refer to as “Practical, Scalable, & Trusted AI.”
Practical & Scalable AI
For most P&C insurance providers, the current claims workflow is a highly manual process. Due to the large number of manual tasks, the end to end process flow is prone to human errors and leads to large time gaps between the event and claim completion_ that all important time to settle KPI. Today insurers utilize current events, location data, and IoT device data to retroactively determine root cause of a claim, but generally, useful insights generated in this process are few and far between.
Utilizing AI, insurers can leverage the same data plus external 3rd party data to:
- generate actionable insights and greater accuracy in predicting and preventing future claims
- reduce fraud due to the addition of risk & data checkpoints along the process
- provide a faster time to settle for their customers
AI practitioners are able to utilize machine learning to identify claims risk and denial probabilities. AI can also prescribe preventative actions prior to a claim event occurring, as well as leverage machine learning & natural language processing to identify potential claim causes.
Let’s take a look at what AI is doing in the auto claims space.
The new AI-powered workflow starts with the customer uploading an image or set of images via the customer’s mobile application. After the upload, AI powers the entire process. Via numerous algorithms, such as those from CognitiveScale and other 3rd party vendors, the AI solution can automatically generate a wide variety of data insights such as car make and model, color, license plate, and weather at the time of the event to help suggest causality. All of this information is then tied back to the insurer’s CRM system to ensure that the car belongs to the proper owner and lower the risk of possible fraud. With AI, there’s also much more an insurer can do:
- A push notification can be sent to the customer & to their closest approved tow truck drivers and body shop locations.
- If the customer submits a picture of the other person’s car (or license or insurance card), you can then determine what insurance company they are using and pull their DMV records via an API call or using a Robotic Process Automation bot to fill out and fax the appropriate forms.
- Most insurance companies don’t have adjusters and municipality in every state, so they hire 3rd party claims adjusters who are requested from a pool. The insurer can immediately push the claim from the mobile app to the closest 3rd party claims adjuster so they can come look at the vehicle.
- The final point is the most potentially valuable to the auto insurance industry_ AI can determine the extent of damage to the car & whether it is a total loss or not.
Just as we saw in the claims workflow, today’s underwriting process is a slow and tedious collection of manual processes. Due to the large number of manual tasks and the lack of readily available information, the end to end process flow is prone to errors and can lead to a more risky portfolio than desired by small & large insurers alike. Today, insurers attempt to utilize a select few data sources to accurately predict risks, and this can lead to missed insights that would otherwise prove useful in the underwriting process.
Through AI, insurers can leverage hyper-personalized profiles to generate insights into new and existing customers. A hyper-personalized profile is created using standard company data along with external 3rd party data to determine the declared, observed, and inferred insights that profiling traditionally misses. With this process, insurers will have more personalized policies, gain greater value out of the available data & reduce overall risk exposure.
So, how exactly does this work? AI solutions can utilize natural language processing to identify latent factors about the underwriting prospect & property. Machine learning is then leveraged to provide recommendations with evidence & explainability, combined with other learning models for risk appropriation. When combined, you get a full view of both the entity being underwritten & hidden risks associated with the potential policy. The following are other examples of how AI is revolutionizing the underwriting process:
- Predict premiums based on past risk assessments to make the risk assessment more precise
- Detect potential fraud based on speech and tone analysis of customer calls
- Utilize computer vision to predict risk and premiums without needing to send out a field underwriting resource
- Suggest risk categories of potential customers based on similarities to existing customers
As these AI systems become more sophisticated and embedded in insurance workflows, the onus falls on both the insurance provider and industry regulators to ensure its responsible design and use. Doing so is required to ensure that this powerful technology is properly applied, and those of us working in AI have an ethical & responsible duty to both insurance customers and the greater community for increased explainability and “data robustness” —the ability to detect adversarial data responses and protect against adversarial attacks.
The insurance industry continues to expand the day to day practical use of AI. As this proliferation continues, one of the most important topics to pay attention to is trust. Trust ensures that we are thoughtfully designing and managing a continuous chain of responsibility and autonomy encompassing the ethical acts and decisions of systems that learn and adapt with use. To help make this happen faster and safer, a cross-industry approach is essential, and that is why CognitiveScale has partnered with numerous cross-industry organizations such as the University of Texas at Austin, USAA, Microsoft, IEEE, & PWC). These partnerships have created an open community called AI Global to build trusted frameworks for numerous industry verticals, including P&C insurance. By partnering with this organization, CognitiveScale is ensuring that the future use of AI in insurance will be transparent, accountable, and trustworthy.
For more information on CognitiveScale and AI Global, please see the links below: