2025 Year in Review: Health Systems Sought to Prioritize Artificial Intelligence Use Cases

One of the challenges for healthcare journalists this year has been to determine where we are on the Gartner Hype Cycle with AI. We have definitely passed the Innovation Trigger, where a breakthrough sparks interest, and seem to be squarely in the Peak of Inflated Expectations, where initial hype can lead to unrealistic hopes. We have not reached the Trough of Disillusionment yet. That is probably coming. 

And because almost every vendor product announcement explains how a solution is now “AI-driven,” it becomes difficult for reporters to tell the wheat from the chaff. At Healthcare Innovation, our approach has been and will continue to be to stay close to the health system CIOs, CMIOs and chief AI officers as they prioritize use cases and solutions and seek to quantify the quality and efficiency gains they are seeing. We also try to interview venture capitalists and consortium leaders zeroing in on AI startups in healthcare. 

This month we published our annual package of 10 Transformative Trends, and this year we devoted the entire package to AI with topics ranging from  governance and regulation to how AI is impacting areas such as oncology, revenue cycle, and cybersecurity. I encourage you to spend some time with that package, but I also would like to take a look back at some of the ways the use cases are evolving, and what some health tech execs and investors told us they expect to happen next. 

In October, venture capital firm Menlo Ventures published a report on AI adoption in healthcare. The research indicated that 22 percent of healthcare organizations are using domain-specific AI tools, a sevenfold increase from 2024 and a tenfold increase from 2023. Health systems have the highest adoption rate at 27 percent, with outpatient providers at 18 percent and payers at 14 percent.

According to Menlo Ventures, the $4.9 trillion healthcare industry, which accounts for one-fifth of the U.S. economy but only 12 percent of software spending, is now adopting AI at more than twice the rate of the overall economy (2.2x). Healthcare AI spending hit $1.4 billion this year, nearly tripling 2024’s investment.

Two categories that address urgent operational pain points and deliver measurable ROI are ambient clinical documentation ($600 million) and coding and billing automation ($450 million). Other fast-growing categories include patient engagement (+20x year over year) and prior authorization (+10x year over year).

In September, I interviewed Murray Brozinsky, a partner at Aegis Ventures, a New York-based venture studio that partners with entrepreneurs and health systems to launch and scale health tech startups. He referred to ebbs and flows of enthusiasm around AI. “Everybody got excited by the zeitgeist moment of ChatGPT and then it went through a little bit of the valley of disillusionment once they started to try to implement these things,” he said. “But it's certainly the No. 1 thing on everyone's mind. How do you apply generative AI  or agentic AI to workforce issues? We're absolutely seeing that as being the No. 1 priority — how to implement to get efficiency and effectiveness. So far, the focus has been more on efficiency. How do I do what I'm doing now but do it better? We feel there's a lot of opportunity there, but we actually think there are problems that you couldn't solve previously that now you can revisit and use AI to solve.”

In June I spoke to Ochsner Health CIO Amy Trainor, R.N., about how the health system was expanding its relationship with a company called Latent to develop clinical AI tools to help with specialty, infusion, and retail pharmacy prior authorizations.

In that conversation, Trainor said something I found intriguing: “What we don't want to do and we don’t want Latent to do either is have a war of bots between the payers’ AI and our AI, and let's see who wins. That's not success for the patient, for the payer, or for us as an organization.”

A few weeks later I had a discussion on this same topic with Fawad Butt, founder and CEO of startup Penguin Ai and former chief data officer of UnitedHealthcare, Kaiser Permanente and Optum. His company says its flagship platform combines task-specific small language models (SLMs) and digital workers and agents with a healthcare-specific AI platform to streamline processes such as prior authorizations, claims processing, medical records summarization, and appeals management.

I asked Butt about this idea of a battle between payers’ and providers’ AI agents.
“That war has started,” he said. “The agent wars are here, right? It is not this futuristic thing that's going to happen. It's happening today. I sat with the CEO of one of the largest regional health plans in the country. He said what they are seeing is that, in some ways, the providers have adopted agents a lot quicker than the payer side, because the payers’ processes are more complex. In one scenario, he said, a small network of providers that used to do 5% appeals on denials is now doing 100% appeals on every denial the health plan is sending them. He believes the provider group has an agent on their side, and the health plan has eight people on its side. So how are they going to win that?”

Monitoring AI performance

In 2025 many health systems continued their work to establish AI governance teams and processes, and we reported on several startups that have been formed to help with those efforts. Among those are Vega Health, led by CEO Mark Sendak, M.D., who previously served as the Population Health & Data Science Lead of the Duke Institute for Health Innovation. His company is partnering with health systems to implement solutions in their local environments and monitor performance objectively. Vega Health also works with health systems and innovators to commercialize effective AI solutions across the healthcare ecosystem.

Another company in this space is Qualified Health. In a June interview, Kedar Mate, M.D., the company’s co-founder and chief medical officer, explained how the company is helping health systems build infrastructure to support generative AI initiatives.

Mate, former CEO of the Institute for Healthcare Improvement, said that the initial approach to governance by most health systems has been to say we have a governance committee for our technologies and now we have a subcommittee focused on generative AI or AI solutions writ large. "We call that analog governance, for an analog era, for analog technologies. But we need truly digital governance,” he stressed. “These AI tools demand digital governance because of their nature. The underlying foundational models evolve over time and that will cause either augmentation or degradation of performance of the specific application that is in use in your institution. And you might not know about that unless you are regularly monitoring the performance of those algorithms.”

He added that one of the company’s premises is that automation is going to be fundamental to how we drive improvements in productivity. “Your market competitiveness will be determined in part by how you adopt AI in the future. If you adopt it sooner and better than the system across the street, then you're going to have a better opportunity to corner aspects of the market in the future,” Mate added.

Going back to my interview with Ochsner’s Amy Trainor, I asked her whether she had a sense that the large, entrenched players such as major EHR vendors were moving fast enough on AI tools.

“Six months ago I would have said no,” she responded. “I don't know what happened there, but it has been like a massive change. With most things, we're going to look at our partners first. If they don't meet our needs, we're going to look elsewhere. It is the same with AI. We want to make sure we're capturing the value of what we're spending already. In general, we're going to look internally first. I would say they are moving a little faster. Latent was a great example —Epic’s not doing this yet, but they probably will be in the next six months.

“Turning on AI is a much different conversation than a color change or a new button on somebody's dashboard. We have to make sure that this is safe and follows our responsible AI principles,” Trainor added. “I would say that the ChatGPT era is the fastest-moving we've ever been able to take advantage of any public tech in my 20 years in health IT.”

Stay tuned to follow more fast-moving AI developments in 2026.