How AI is moving health-care costs in the wrong direction

Although artificial intelligence was expected to reduce costs in health care, emerging evidence suggests that it may be increasing spending for many payers.

Rob Duffy, chief technology officer at HealthEdge, explained how AI is increasing costs for many payers.

First, Duffy said, there is the medical loss ratio (MLR) bucket, which is the actual cost of care, and that’s where AI is creating the most obvious upward pressure right now. Then, he added, there’s the administrative loss ratio (ALR) bucket, which is the cost to own and operate the plan, claims operations, systems, labor, integrations, and other functions.

“The reason both matter is that AI can push costs up in both places: externally, through more utilization and more optimized claims, and internally, if plans respond to AI with more complexity instead of more scale,” he explained.

The first place you see it is in utilization, Duffy said. Consumers are getting smarter with their health care, and tools like Chat GPT mean that people are self-diagnosing, researching symptoms and going to see doctors in situations where they may not have gone before. Some of that is good, he pointed out. After all, “we want members to be engaged.”

But from a payer perspective, it means more visits, more services, more claims, and more utilization flowing through the system, Duffy added. That shows up directly in MLR.

'That makes complete sense'

The second big area is the provider revenue cycle, Duffy said. Providers are using AI to code claims more effectively, capture documentation more completely, and optimize reimbursement.

“Again, from the provider’s side, that makes complete sense,” he said. “But from the payer side, it can increase the amount paid on claims and create even more upward pressure on MLR. So, yes, AI is increasing the volume of care, but it’s also making the claims that come through more economically optimized for the provider.”

And then there’s a third cost risk, which is operational and where ALR factors in, Duffy added.

If plans just bolt AI tools onto fragmented legacy systems, they can actually increase their administrative loss ratio instead of reducing it.

“You get more integrations, more data movement, more reconciliation, more vendor sprawl, and more one-off proofs of concept that don’t scale,” Duffy said. “So even though the biggest AI-driven cost pressure is showing up in MLR, payers also have to be careful that their AI strategy does not create a new layer of ALR cost on top of an already complex operating model.”

Reducing costs

So, what are some of the steps that payers can take to reduce these increasing
costs? The first thing that they need to do is to be very clear-eyed about what they can and cannot control, Duffy said.

They are not going to stop consumers from using AI to understand their symptoms, and they are not going to stop providers from using AI to code more effectively and optimize the revenue cycle.

“Over time,” Duffy said, “plans absolutely need strategies for both of those things, like better member education and guidance around AI, and better payer-side AI in areas like payment integrity, claims review, and utilization management,” he added.

But where a plan can get more immediate relief, and where it has the most direct control is the administrative loss ratio, Duffy said.

“The question I’d be asking as a business leader at a plan is: Can we accelerate ALR savings fast enough to offset the MLR pressure that is already here?" he added. "Reducing administrative costs through automation, scalable platforms, and AI that is actually embedded into the operating model – is the lever payers can pull right now."

That starts with infrastructure.

“If you have different lines of business, different regions, different versions of core admin platforms, and hundreds or thousands of integrations, you are not going to get the leverage you need from AI,” Duffy said. “You have to simplify. You have to consolidate. You need a single view of the data and a platform that lets you operate at scale, because otherwise, every AI use case becomes another custom integration project.”

Stop thinking of AI as a bolt-on

The second step is to stop thinking about AI as a bolt-on and start embedding it into the operating model of the plan, Duffy said. AI needs to be deeply connected to the system of record so it can consume information, trigger workflows, support agents, reduce manual clicking, and actually change the cost structure of the business.

“That is where you start to get real ROI, from AI that is integrated into claims, care management, payment integrity, member engagement, and the core administrative workflow,” he said.

Finally, payers need to decide if they can build and maintain that capability internally fast enough, Duffy said. Because the market is moving very quickly.

“If you can build the skill set, ship rapidly, govern it properly, and keep driving ALR down, great,” he said. “But if not, you should be asking whether there is a partner who can guarantee outcomes, continue investing in the technology, and help you move fast enough so your business stays healthy while your core workforce focuses on member engagement, health outcomes, and taking care of the member population."

Ayo Mseka has than 30 years of experience reporting on the financial-services industry. She formerly served as Editor-In-Chief of NAIFA’s Advisor Today magazine. Contact her at amseka@INNfeedback.com.

Ayo Mseka has more than 30 years of experience reporting on the financial services industry. She formerly served as editor-in-chief of NAIFA’s Advisor Today magazine. Contact her at amseka@INNfeedback.com.