Bot vs. Bot: Why Healthcare AI Progress Might Be Stuck

Healthcare is in the middle of an AI gold rush, with investors, providers and payers pouring more money than ever before into new technology. But according to a group of top healthcare CFOs and industry leaders gathered at HFMA’s annual conference, much of that investment is destined to disappoint. 

The problem isn’t necessarily the AI tools themselves — it’s more that most healthcare organizations are bolting these sophisticated models onto infrastructure that is decades-old. 

Seema Verma — former CMS administrator and general manager of Oracle Health & Life Sciences — put it bluntly: “You can’t have an AI strategy without having a data strategy.” 

She argued the infrastructure most providers are working with today is simply too antiquated to support what AI actually requires — namely, access to real-time, unified data across clinical, financial and operational systems. 

For instance, a physician relying on an AI recommendation needs that model to know whether a patient’s health plan will cover a medication, whether it’s in stock, and whether the patient has any interactions — all in the moment. 

Without that connective tissue, Verma said even the strongest AI tools will underperform.

Mike Marks — CFO of HCA Healthcare, one of the largest health systems in the country — echoed Verma’s concerns and added that the sheer scale of technical debt is actively blocking progress. 

“The amount of legacy systems that we’re all dealing with is really getting in the way of transformation,” he remarked.

It’s incredibly expensive to replace technology at that scale, and the demand for AI innovation already far outpaces any health system’s ability to pay for it, which makes getting the foundational work right even more critical, Marks noted.

His formula for prioritization is: clinical systems first, then operations and administrative functions. 

The logic is patient-centered. Clinical systems touch care delivery directly, so improvements there have the most immediate impact on health outcomes, Marks explained. Operations and administrative functions matter too, but in Marks’ view, they should be optimized in service of a system that is already delivering high-quality care — rather than the other way around.

Clinical AI that underperforms has direct consequences for patients, so it demands the most rigorous investment and oversight, Marks added.

Another health system CFO — Scott Hawig of BJC Healthcare — agreed that the healthcare industry still has a long way to go before AI can deliver on its promise.

The current reality, Hawig argued, is best captured by what’s happening between payers and providers. He painted it like this: two sets of bots, each deployed by opposing sides, endlessly fighting over claims with no one winning. 

“Bot versus bot — provider revenue cycle bot against the denial insurance bot — is the fundamental problem,” Hawig declared.

The AI tools exist, the investment capital is flowing, and the urgency is real. But as these experts made clear, writing huge checks without fixing the data foundation first might just be a more expensive way of staying stuck.

Photo: imaginima, Getty Images