Report by HFMA and Eliciting Insights examines hospitals’ journey to realize the potential of AI

Although healthcare AI technology has been in the pipeline for years, new feedback suggests implementation remains a work in progress at hospitals and health systems.

As described in an HFMA report produced with Eliciting Insights, most organizations are still striving to develop an AI strategy and install an AI governance structure.

Findings of the report (available for purchase) show that 88% of organizations use AI in some form, and 71% have identified and deployed pilot or full solutions in finance, revenue cycle management or clinical functional areas.

Yet only 18% have a mature AI program, defined as having both a later-stage governance structure and an evolving or complete AI strategy. That select group of health systems generally is larger in net patient revenue and bed count, although more than a quarter in the category have fewer than 900 beds.

Meanwhile, 38% of respondents said they have a comprehensive AI strategy but not an established governance structure, a particular challenge in healthcare amid intensive regulatory considerations and the scattered nature of patient data. And 42% have not developed a true AI strategy.

Even among health systems with mature AI programs, only half report having sufficient resources to implement AI solutions. In the overall survey pool, more than 80% have a dearth of such resources.

Those insights were some of the key takeaways from survey responses of 233 health system leaders and qualitative interviews with CFOs during 2025 Q2.

A tricky dynamic

The challenge of AI implementation reflects the continuous pressure on hospitals to balance the adoption of new technologies with cost management and quality improvement, said Todd Nelson, FHFMA, MBA, chief partnership executive and director of partner relationships for HFMA.

“To that end, they are reviewing the promise of efficiencies that can be gained by adopting technology such as AI, but with a keen eye to the impact on operations, quality and cost — and the determination of risk to the business,” Nelson said.

That risk includes the increased potential for misalignment among platforms.

“A greater number of ‘handoffs’ between multiple systems makes it riskier to the business as you have multiple configurations, contracts and relationships to manage,” Nelson said.

There is a learning curve with respect to AI risk assessment, survey results suggest.

“While AI is transformative, many health systems are struggling with understanding and safeguarding against the risks associated with using AI in healthcare,” the report states, pointing to issues including not only data security but also patient privacy, ethical concerns, adherence to regulations and patient safety.

Impetus for investment

Even as cost reduction drives AI investment, only 39% of responding CFOs expect their investments to reduce costs. Conversely, the cost of AI solutions is viewed as a top barrier. Data-sharing issues then become a notable roadblock after organizations begin to invest in AI.

Six in 10 health system CFOs believe revenue cycle represents the best opportunity for AI, according to the survey, reflecting an increasingly heard theme within the industry.

AI is “a kind of impetus for solving a lot of the friction that exists in healthcare,” Judson Ivy, founder and CEO of Ensemble Health Partners, said in May during an AI-focused session at the annual Not-for-Profit Healthcare Investor Conference. “You still have to figure out what the friction in the ecosystem is that causes this administrative waste. And so much of that friction is around payer-provider [interactions] and really non-value-added work that occurs.”

Even amid a focus on revenue cycle applications, the HFMA survey finds that ambient listening in clinical settings is one of the top AI solutions being adopted, along with clinical documentation improvement.

“I look at AI as a capability that makes time and space for doctors,” Allon Bloch, co-founder and CEO of the primary care company K Health, said during the investor conference. “Doctors can be much more efficient. They can also manage patients very differently in the visit and after the visit.”

Crucial relationships

Of health systems that are still working to develop a mature AI program, per the HFMA survey, 70% rely on vendors to provide guidance on AI opportunities. There appears to be an increasing eagerness for solutions, according to industry feedback.

“We’ve essentially broken the healthcare sales cycle,” Harpreet Mangat, chief strategy officer with Hippocratic AI, said during the investor conference. “Educating people on a new technology, getting to building trust, building credibility, showing them proof points — normally that’s a game that takes 18 months to two years. We are coming back sometimes with signed contracts within two weeks, deployed within a few months.”

Such an outcome is more likely when providers are familiar with the vendor, according to the survey findings. Most health systems are more comfortable piloting solutions and sharing data with their existing vendors or with AI companies that partner with the organization’s vendors. Roughly one in 10 plan to wait for a solution from their electronic health record (EHR) vendor.

Hospitals are “focusing on leveraging existing relationships to their fullest potential before looking at developing new ones,” Nelson said, noting that investment in multiple interfaces generally means higher costs.

Said Ivy, “There are a lot of questions about how all this is going to work together, and is it a wait-and-see to say, ‘Is there going to be an integration layer? Do we wait for the EMR to do these things?’ … I think it’s a question of, ‘How do we not end up in a [situation] where we’ve got 60 companies we’re working with that have some AI or agentic technology that [are] not communicating with each other, and then we’re creating more friction?’”

About the report

The HFMA and Eliciting Insights report provides actionable insights on health system readiness for AI, a deep dive on health systems with mature AI programs, current and planned areas of investment, expectations for vendor performance and barriers to adoption. It also includes four persona profiles that illustrate the different decision-making dynamics within health systems. HFMA Peer Review customers receive the report as a benefit.