There’s a lot of talk about artificial intelligence (AI) in insurance, but how exactly are insurers leveraging new AI and machine learning (ML) capabilities? Whereas some of the hype may seem nebulous, specific use cases are emerging. Here are six ways insurers are using AI/ML in underwriting.

1. Supporting Human Underwriters

Underwriters spend 40% of their time on noncore activities. Accenture estimates that insurers will lose $85 billion to $160 billion over the next five years in inefficiency as a result. AI can reduce this waste.

Using a modern AI program is like having a personal assistant to summarize documents, answer questions, and draft emails. By utilizing AI tools, underwriters can eliminate some of their most monotonous tasks, which allows them to focus on high-value activities that require a human touch.

2. Selecting Risk

Risk selection is at the heart of what underwriters do. AI can help them do it better.

Insurers often turn off entire segments associated with unacceptable levels of risks. Whereas this may help them avoid losses, it may also lead them to deny policyholders that would be profitable. With AI, this is no longer a necessity.

Let’s say an auto insurer has identified a high rate of losses among drivers with very low credit scores. To avoid these losses, the insurer turns off this segment. As a result, the insurer misses out on some safe drivers who happen to have low credit scores.

Underwriters may see this as necessary because they can only consider so many factors. With ML-powered algorithms, they are no longer constrained. They can consider a multitude of additional factors to get a more accurate picture of each risk’s rate adequacy. This enables insurers to take on more business while more carefully considering risks that are potentially inadequately-rated. 

3. Detecting Fraud

Insurance fraud is a major cost for insurers, but AI can help reduce it.

In a survey from ValuePenguin, 21% of auto insurance customers admitted to having misled their insurers to save money. Garage fraud is one of the most common types of deceit – 16% of Millennials said they’ve used another address to secure a better rate. Other common types of application fraud include leaving out drivers and underreporting how much the vehicle is used. AI and ML can help underwriters catch these misrepresentations. ML excels at combing large databases and spotting patterns that tend to indicate discrepancies – something that is hard for human underwriters. If the program spots indicators of a discrepancy, it can flag it for review.

4. Fine-Tuning Rates

Before insurers can file new rates, they need to pour over mountains of data. AI programs and workflows make this process exponentially faster and more accurate.

Furthermore, a risk scoring software can provide an early indicator of the need for rate adjustments. When the scoring software detects trends in the percentage of risks that are rate adequate or the percentage of quotes that convert to bound policies, the insurer can quickly adjust their rate filings.

5. Improving Customer Experience

When drivers request auto insurance quotes online, they don’t want to wait days. In many cases, the insurer that replies the fastest is the one to make the sale.

AI can reduce wait times. Young consumers, in particular, are comfortable with AI – as long as it improves the customer experience. In a Policygenius survey, 43% of consumers between the ages of 18 and 34 said they would trust AI to process their application accurately and faster than a human.

6. Reducing Bias

Bias is often cited as a risk to avoid when implementing AI/ML programs. Whereas it’s true that some AI/ML outputs are biased, bias is not an inherent characteristic of AI/ML.

In fact, research published in the European Journal of Marketing found that the use of AI/ML can even counter human bias. This stems from the technology’s ability to weigh multiple factors instead of relying on broad categories. For example, applying higher prices in low-income ZIP codes may be discriminatory. AI/ML programs can facilitate an individualized risk selection approach.