Is integrating AI and automation in health insurance worth the risk?

Cumbersome paperwork and manual processes have burdened health insurers for decades, but the industry is finally ready to embrace new technologies to improve efficiency and client satisfaction. 

Unfortunately, it hasn’t been an entirely smooth transition to AI automation in health insurance. Recently the families of two deceased UnitedHealth policy holders are suing after the company allegedly used a faulty AI that denied elderly patients coverage for medically necessary care. The American Medical Association also warned of potential risks for health insurance companies using AI and is now advocating for greater regulatory oversight. 

So is integrating AI and automation in healthcare worth the risk? Or do health insurers need to take a step back until we’ve solved some of the problems inherent in AI systems? 

AI in health insurance

Many insurers are already using AI, although adoption has been fairly recent. In 2017 only a little more than 1% of insurers were using AI, compared to 30% in many other industries. Today, experts estimate that AI in insurance will be worth $35.77 billion by 2030, with a compound annual growth rate of more than 30%. 

AI helps insurers accelerate underwriting and claims processing by analyzing historical data to evaluate risk. Insurers are also using AI to detect fraud, to generate accurate loss reports, and more. As a result, the effective use of AI can improve claims accuracy by up to 99% and increase efficiency by around 60%. That means customers get policies settled and claims approved faster, resulting in a greatly improved customer experience. 

Of course, there are also the more visible uses of AI in health insurance. If you’ve ever visited an insurer’s website, you’ve probably interacted with a chatbot. And chances are that chatbot didn’t understand your questions or was unable to give you the answers you needed. But now with software like ChatGPT and Bard helping chatbots become more understanding of the needs and complaints of customers, insurers can offer more personalized assistance on their websites. A chatbot could even guide a customer through the steps of submitting a claim, including what documentation the customer needs to include. 

In short, AI can help health care providers better serve their customers and enhance an experience that’s too often clunky and complicated. 

Pitfalls of AI in insurance

So now you’ve seen the good side of AI in insurance. But we can’t forget about the bad side. The AI model that UnitedHealth deployed may have had up to a 90% error rate according to the lawsuit, and if you end up using AI in the wrong way, you could find yourself facing a similar lawsuit or causing harm to your customers. Basically, you end up with a defective AI system when the data you train it on is wrong. Maybe someone tampered with the data, or there just wasn’t enough data to make the right decision. 

And then there’s bias, something AI systems are often accused of. For instance, let’s look at age bias. Let’s say, for example, your AI system needs to examine a claim from a customer trying to get coverage for a health care procedure that typically only patients of a particular age require. If the AI system determines, based on the data it has available, that the procedure isn’t required for people under the age of 40, your system might reject the claim or consider the procedure optional when in fact it’s absolutely necessary. Insurance regulators are working on finding ways to address potential bias in insurance models, so you should carefully test your systems for bias. 

It’s important to train your systems the right way, but ultimately you need to have humans involved in the final decision, particularly in matters of life and death. Even the best AI systems make mistakes, exhibit biases, or simply make things up. That’s why you shouldn’t use AI to replace humans in health insurance; AI should complement the work that humans are doing and make it easier for your employees to do their jobs. This will be especially important as regulatory oversight increases. 

The future of AI in insurance

Many insurers are already using AI, and that should continue. Artificial intelligence is an essential technology that helps to ensure customers get timely approval for claims and can receive the coverage they need. 

But health insurers need to be careful that the AI systems they use create efficiency without introducing costly and unfortunate errors. Typically, this requires that people remain at the center of the underwriting process, providing the judgment and empathy that, at least at this point, AI simply can’t provide.