Even AI optimists don't envision the technology fundamentally remaking the U.S. health care system anytime soon, but there's widespread agreement that it has the potential to vastly improve the quality of care and trim costly waste.

Why it matters: The scale of change that AI could bring to health care not only impacts patients but also the millions of people the system employs — who will ultimately shape how widely it's adopted.

The big picture: Recent breakthroughs in AI technology are coming up against a health care system that is very resistant to change, in no small part because of how heavily it's regulated and the trillions of dollars at stake.

• That will temper AI's adoption — especially in ways that cost jobs or money. But on the other hand, uses that drive up revenue, increase productivity or improve health care workers' quality of life will be attractive and therefore more likely to be integrated into the system.
• For example, there's a lot of talk about AI's potential within biotech to discover more drugs, speed up the development process and better match drugs to patients.
• Other commonly discussed uses include improving diagnostics, reducing providers' paperwork hassle and streamlining billing processes.

"A lot of the transformation will occur by helping the companies that we fund be better at what they do and more efficient. So it's the efficiency that is probably key," said Dan Mendelson, CEO of Morgan Health, a unit of JPMorgan Chase.

• "I don't have any hopes that within the next three to five years that these technologies are going to be transformative in terms of patient interaction with the clinician," Mendelson added. "Dr. Google is still Dr. Google."
• "My vision for AI is that AI is going to do what computers do really well, and it's going to allow humans to do what humans do really well," said Rafael Rosengarten, a member of the board of directors of the Alliance for Artificial Intelligence in Healthcare and the CEO of precision medicine firm Genialis.

State of play: Emerging research is illuminating the promise of AI, finding that it can be just as good or better than human doctors at talking to and diagnosing patients or reducing preventable hospital deaths, for example.

• Other research has highlighted its limitations, and reporting has found it's already being used in ways that can be harmful to patients.
• Although no drugs that were discovered or developed using AI are yet on the market, a recent TD Cowen analysis found that there are at least dozens of candidates that were developed using AI in phase 1 or phase 2 clinical trials.

What they're saying: "There's an opportunity to dramatically improve the way we deliver health care either by augmenting or automating in some way," said Suchi Saria, CEO of Bayesian Health and director of AI and health at Johns Hopkins.

• "Health care is definitely a much more complicated sector than many other sectors where we've been successful with tech, so we have to come at it with much more humility and deeper understanding," she added.

Between the lines: Just how ingrained AI becomes in health care will depend on the incentives in place.

• In some cases, it already makes sense for AI to be used, and AI can ultimately create a scenario that benefits patients and various industry players, Saria said.
• But in others, adoption will be resisted by those who stand to lose money or employment from it — be it front office staff, lab techs suddenly made redundant or providers who would lose revenue from performing fewer procedures in some cases.
• And in still others, the right incentives may be there but the urgency isn't — something government could help with, she said.
• "It's very hard to imagine us succeeding if the incentives don't support it," Saria said.

What we're watching: Some AI visionaries see the technology fueling more revolutionary changes down the road.

• "What has happened in all these other industries from the chip industry, to the automotive industry, the same thing will happen in our industry," said Chris Gibson, cofounder and CEO of Recursion Pharmaceuticals, a "techbio" company that uses AI to enhance drug discovery and development. "It's a question simply of when."
• "Maybe we're going to actually understand biology well enough to prevent disease earlier," he added. "Twenty years from now, hopefully we can prevent the majority of diseases and then when people do get them, treat them better."
• In the 5-year to 10-year range, "people are going to have the courage to take the doctor out of the middle and have these tools interface directly with patients to deliver care," former FDA commissioner Scott Gottlieb said at an event hosted by Recursion and Nvidia during the week of JPM's health care conference.

The bottom line: "When you have everybody doing these little incremental adoptions, eventually there's gonna be an inflection point," said Rosengarten, the Genialis CEO. It'll "be like flipping on a switch, because all of a sudden, the entire industry will be driven by AI-based technologies."