Is AI a bandaid for healthcare’s historic problems rather than a fix? A former insurance exec weighs in

Craig Samitt’s healthcare experience is diverse, spanning both the clinical and insurance side of the industry.

A primary care physician by training, he eventually rose to EVP, chief clinical officer and diversified business group president at insurer Anthem (now Elevance Health) in 2015 and became president and CEO of Blue Cross and Blue Shield of Minnesota in 2018, which has over 2.5 million members. He retired from Blue Cross in 2021, and now works as a managing director for ITO Advisors, a healthcare consulting and venture capital firm. 

Samitt spoke with Healthcare Brew about how he would redesign healthcare, and the tension points of AI between hospitals and payers.

This interview has been edited for length and clarity.

How are health plans using AI?

Plans are applying AI to utilization management and prior authorization, member services, digital outreach, care management, and then a bunch of administrative functions. That’s valuable, just like in our day-to-day lives: If something takes us hours today and AI can help us accomplish it in a fraction of the time, we should do that. Most of the ways that health plans are using AI today, I would describe as productivity enhancement or workflow optimization. It’s not transforming the way that health plans work. It’s essentially making the way we’ve always worked at health plans faster and perhaps cheaper.

The more interesting question, from my point of view, is not, “How do I make the process faster?” It’s really more about why I have the process in the first place, and if I were redesigning it today from scratch, would I build it the same way? In many respects, health plans and providers are both layering AI on top of a historical—and I would argue broken—relationship process, either within your organization or between your organizations. We should use AI to redesign processes, not just make broken processes faster.

What does your ideal redesign of healthcare look like?

If I were to redesign the process, I’d start by putting the patient first. If we were to create a healthcare system, would we create something that looks like it looks today? I would say absolutely not. Both on the insurance side as well as on the care delivery side. But if we essentially said, “Let’s map the patient journey. Let’s map the member journey. Let’s understand where friction exists. Where are we creating handoffs? Where are we not providing good service? Where are we limiting quality outcomes?” The answer: If I were to have my wish, I would say providers and payers should get together in the same room and talk about the patient journey and say, ”How would we automate, redesign, and in some respects, eliminate steps that are just not serving the patient, and build a healthcare system around that?”

What are your thoughts on if AI makes care less efficient and more expensive?

I’d take your question in two parts. One is, as we think about the use of AI, we should be thinking of it from a perspective of a net benefit to our industry, not a net cost to our industry. I’ve always wanted our healthcare industry to be higher quality at a lower cost. If AI creates an environment where things are easier, faster, outcomes are better, then ideally we should also simplify, reduce unnecessary costs, and make our industry collectively more effective and efficient. So, on the one hand, I would say our aim should not be to drive up the costs of healthcare by making it easier. It should both drive the cost down as well as make it easier.

The second thing that I would say, though, is we shouldn’t make things easier so that we lose sight of what is right for the patient. More specifically, I’d say you could take friction out of the system, but we still don’t want patients to have surgical procedures that are not warranted by the medical evidence. We certainly don’t want patients to be prescribed medications when there is a better, lower-cost alternative. We don’t want people to access fraudulent services. So we don’t want a no-friction system; we don’t want it to be so easy that it becomes chaotic. We want to have a rational system through the use of technology to help support it.

At times, it seems as though AI tools used by health plans and hospitals are in conflict with each other, like over approving a claim or prior authorization. How does that get resolved?

When we look at AI providers and payers, there are certainly areas where [they both] are using AI to optimize their business model. From my point of view, if providers are using AI, for example, to create more paperwork, and payers are using AI to process more paperwork, then nobody’s winning. The better question is whether AI allows us to remove the transaction altogether. One of my questions is: How do we use AI to make the relationship between providers and payers more seamless? There’s obviously a lot of tension around prior authorization, and it’s hard to ask me about prior authorization because I see the merits of prior authorization from both the provider and the payer lens. I look at prior authorization as a way to assure that patients are receiving care that is proven appropriate and evidence-based, but if instead of using a rules-based system for prior authorization, what if we had AI solutions that would guide providers about appropriateness right from the outset? If AI was used as a way to bring immediate information to providers about clinical appropriateness, if there were an automation process, not for approval of services, but for decision support at the provider level, that innovation reduces the contentious relationship between payers and providers, and I think we need more of something like that.