The Hidden Cost of Healthcare AI: Why Premium Prices Don’t Equal Premium Results

I has made an undeniable impact throughout the healthcare industry: from a nurse navigator using an AI triage assistant to prioritize cases and respond faster, to tools to help providers document visits, to systems that automate repetitive, resource-intensive administrative tasks. As the use cases have added up, so too has the investment in AI. But healthcare systems are starting to look more closely at whether they are actually getting their money’s worth.

Many organizations are paying premium prices for AI tools that save minimal time without considering the true cost-benefit ratio. Many AI tools that health systems purchase for thousands per user annually save only five minutes daily, making the return on investment difficult to justify. Organizations are paying premium prices for marginal time savings.

There’s a fundamental disconnect between what organizations pay for current AI tools, the limited value they deliver, and the more capable systems that will emerge in the coming years. Organizations are adopting AI because it sounds like something they should try, but the technology is still in its infancy. Many AI companies today won’t survive because when renewal time comes, they won’t be able to prove their value. As a result, organizations are hesitating and showing signs of fatigue when it comes to certain AI vendors.

According to the 2024 Healthcare IT Spending Report from Bain & Company and KLAS Research, almost half of healthcare providers cite cost as the biggest pain point with their current tech stack. This means that AI tools with high price tags that demonstrate limited ROI only add fuel to the fire.

How did we get here? 

Sometimes, AI tools work well in isolation during the pilot phase, but difficulties often arise when pushed out across an entire health system. This is particularly true when those systems require integration with complex workflows. In fact, some vendors claim that large customers are using their applications when, in reality, they might only be used by one researcher in a single department. This means that these tools haven’t yet been proven and fine-tuned to work seamlessly at scale.

The best advice for health systems today is to ask incumbent vendors about their AI strategy before choosing a new company offering just one solution. Established companies that can integrate AI into existing workflows will outlast point solutions that may not survive the market correction.

The same report found that regulatory and legal considerations are the top barriers to implementing generative AI (38–43% of respondents). These complexities can create additional hurdles that slow adoption. 

Meanwhile, many AI tools are built on public models like ChatGPT, then lightly customized and sold with healthcare branding. The application may sound innovative, but the actual lift they provide is often small. 

To date, billions of dollars have already been spent on healthcare AI, with billions more coming. Yet it’s still unclear how much of that investment actually translates into better care or meaningful time savings. The World Economic Forum observed, “It’s too early to take a position on whether generative AI in healthcare will help, harm or simply squander billions of dollars with no improvement in people’s lives.”

Four questions before your next AI purchase

While the promise of AI is exciting, it’s still early. Many of the tools coming to market simply aren’t ready for prime time. This uncertainty is forcing health systems leaders to reevaluate their investments.

Then there’s the strain of managing multiple point solutions. Maybe one tool handles documentation, another only handles billing and a patient follow-up requires yet another separate tool. The cost and complexity of these point solutions add up over time. Many CIOs are now spending as much time untangling integrations as they are evaluating new technology.

This is one of the reasons health systems are shifting focus. Instead of buying from new vendors, they’re going back to their core platforms and asking how AI is being integrated into the systems they already use. These solutions may not get the same spotlight as the newest startups, but they often deliver a more reliable path with less disruption.

Dr. Daniel Yang at Kaiser Permanente is taking a thoughtful approach to this very issue. The organization is applying system-wide governance to AI efforts across research, clinical operations, education and administration. He believes AI should enhance clinician judgment instead of replacing it. When Kaiser rolled out a generative AI tool, it came with oversight and intentional design.

For all of these reasons, it’s a good time for health systems to step back and ask some foundational questions instead of rushing into their next AI purchase:

1. What problem does this tool actually solve? Look for tools that address specific operational bottlenecks with measurable outcomes. Before any pilot, establish baseline metrics for the problem you’re trying to solve. A good AI tool should improve both efficiency and quality. If it only automates existing processes without improving patient outcomes or staff satisfaction, it’s likely not worth the investment.

2. How much time and money does it realistically save? Calculate the true cost per minute saved, including implementation, training and ongoing support. If you’re spending more than $1,000 per user annually to save less than 15 minutes per day, the ROI likely won’t justify renewal costs. Focus on tools that eliminate entire workflow steps rather than just speeding up existing ones.

3. Is this a pilot, or is it proven to scale? Demand evidence of successful implementation across at least three different organizational sizes and settings. Look for tools that provide consistent results across broad patient populations, and always test in multiple clinical environments before committing to organization-wide deployment.

4. Will this fit into our existing system, or are we adding yet another layer to an already overloaded tech stack? Prioritize tools that integrate directly with your EHR and reduce the number of systems your staff needs to navigate. Any AI solution that requires additional data entry, separate logins or workflow disruptions should be viewed skeptically.

It’s time for healthcare systems to take a more realistic approach to measuring the ROI of AI investments. One that separates true value from the limited use cases that just look good in a demo.

Photo: phive2015, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.