A few weeks ago, I found myself somewhere past 100 kilometers into The Rift, a 140-kilometer gravel cycling race through Iceland, wondering exactly why I had thought this was a good idea.
The scenery was spectacular. We rode through black volcanic landscapes, past mountains and rivers, across terrain that at times appeared so stunning, it looked fake. I felt incredibly lucky just to be there and to have the opportunity to participate in an event like this. I was also exhausted. By that point in the race, the challenge was no longer simply physical. My legs were tired, everything hurt, and I still had a long way to go. Finishing became as much a mental exercise as a physical one: keep moving, keep eating, keep drinking, keep turning the pedals.
Completing those 140 kilometers (about 87 miles) was a huge accomplishment for me, precisely because it was hard. I would not have felt the same way crossing the finish line if someone had removed the climbs, smoothed out the gravel, and shortened the course by half.
On this bike race we endured a lot of friction. It was not a smooth paved flat path. Riders tackled a continuous gravel path, lots of climbs, and a long 87-mile journey across just one day.
Real physician voices, twice a week
Free, and one click to unsubscribe.
Since coming back home from that adventure, I have found myself thinking about that experience in an unexpected context: health care. We talk a lot about friction in health care, usually as something we want to eliminate. But friction itself is not always bad. In medicine, some of it is important. A complex patient should make us stop and think. A physician who disagrees with another physician may cause us to reconsider an assumption. A difficult clinical decision often deserves discussion and deliberation. That kind of friction can make health care better.
But we have also built an extraordinary amount of friction into health care that accomplishes nothing of the sort. Anyone who has practiced in a hospital knows the routine. A physician believes a patient requires inpatient care. The health system sends clinical information to the payer; the payer reviews the case and reaches a different conclusion. Additional records are requested. Nurses spend time gathering back-up documentation. Physicians prepare for peer-to-peer conversations. Cases are denied and appealed, sometimes through multiple rounds of review.
This type of friction in health care is tedious, takes hours and hours of manual work, and detracts from our mission: delivering the right care to the right patient at the right time.
There are certainly situations where that process reflects a legitimate disagreement about medical necessity. Medicine is complicated, and two thoughtful clinicians can look at the same patient and reasonably reach different conclusions. But after spending years working at the intersection of clinical medicine, utilization management, and data science, I have become convinced that a surprising amount of this conflict is not really about a difference in clinical judgment.
It is about a difference in information.
The provider and the payer are often looking at different versions of the same patient. The information may have been captured at different times, come from different parts of the medical record, or been reviewed before the patient’s clinical picture fully evolved. The treating team may have watched the patient change over six or twelve hours while the payer is making a determination from an earlier snapshot. When that happens, we spend an enormous amount of time debating whose interpretation is correct when the real problem is that the two sides did not start with the same picture.
That distinction matters even more as AI becomes part of utilization management. Today, payers, on their side, are investing in AI to make reviews faster and more efficient, while providers, on their side, are investing in AI to identify denials, generate appeals, and prepare physicians for peer-to-peer conversations. Many of these tools solve real problems. As a physician who has spent years watching clinicians buried in this back-and-forth administrative work, I am certainly not arguing against technology that can give some of that time back.
But these AI investments are fueling the payer-provider debate versus uniting the discussion around shared, real-time data to deliver better patient care. Instead of asking how technology can help one side become better at the utilization management argument, we must focus on where technology can create a shared, objective view of the patient before the argument begins.
I am the chief architect behind the Care Level Score, an AI model that uses clinical information from electronic health record (EHR) systems to create a real-time measure of a patient’s clinical complexity and anticipated level of care needs. Importantly, it’s dynamic and changes as the patient’s condition changes. The score is not intended to replace physician judgment or make an admission decision. Its value is in giving the provider and payer a common clinical signal based on the same current information.
Instead of a provider sending records and a payer reviewing a static snapshot through an entirely separate process, both organizations can begin with a shared understanding of a patient’s clinical picture. Through an approach we call Precision Utilization Management, providers and payers can agree in advance on how that shared signal is incorporated into their utilization management process. When the clinical picture strongly supports inpatient care, those cases can move through the system with far less manual intervention. The payer maintains oversight and the physician maintains clinical judgment, but neither side has to spend hours proving something the data already makes clear.
To me, that is where AI becomes much more interesting. The real opportunity is not to automate every decision or remove humans from the process. It is to use technology to distinguish between the cases that genuinely need human judgment and the cases that have historically consumed human time simply because our systems did not support efficient workflows and communication.
While this approach streamlines the process, we should still absolutely spend time on difficult cases. If two clinicians look at the same patient, with the same information, and disagree about the appropriate level of care, that conversation has value. It may change someone’s mind. It may uncover something that was missed.
That is meaningful friction in health care.
Repeatedly sending medical records, reconstructing a clinical picture that already exists and appealing a decision that originated from incomplete information is a different kind of friction. It consumes physician and nurse time, creates expense for both payers and providers, and ultimately adds another layer of complexity to the patient’s experience without making the clinical decision better.
AI gives us an opportunity to reduce this undesired friction. It also gives us an opportunity to institutionalize it at a speed and scale we have never seen before.
AI belongs in utilization management. In fact, it’s already here, uniting hundreds of hospitals and health systems with their payer partners to streamline the process and deliver better care to patients. As we advance AI in health care, the most important question is what problem we are asking AI to solve?
Are we using it to make each side better at navigating an increasingly complicated system? Or can we use it to remove complexity that never needed to exist?
As I think about these questions, I filter them first as a physician. The goal is the same one I learned when I was taking care of patients in the hospital: the right patient, in the right place, at the right time. The technology should help us get there with greater confidence and less unnecessary work, not simply make an inefficient process run faster.
I would not remove the hard parts from those 140 kilometers in Iceland. The climbs, the gravel, the exhaustion, and even those moments when I questioned whether I could finish are part of what made crossing the finish line meaningful.
But health care is not an endurance race. We do not get extra credit for making the journey harder.
Some friction makes us better. We should work hard to remove the rest.
This essay is cited in the KevinMD record on prior authorization.
Heather Bassett is the chief medical officer of Xsolis, where she has served since 2013, and she brings more than twenty years of health care experience. She pioneered the Care Level Score, an artificial intelligence (AI) model that delivers a real-time measure of clinical complexity for utilization management. Her experience as a hospitalist shaped her focus on reducing the administrative burden that pulls clinicians away from patients.
She earned her medical degree from the University of Texas Medical Branch and completed her internal medicine residency at Vanderbilt University Medical Center. She is board-certified in internal medicine, is affiliated with Centennial Medical Center in Nashville, and is active in the Coalition for Health AI.
Her writing on responsible AI adoption, health technology trends, and clinical workflow has appeared in Forbes, HIT Consultant, MedCity News, and Managed Healthcare Executive. Her work has also been featured in Newsweek and Healthcare IT News. She hosts AI Amplified, a podcast syndicated on Healthcare NOW Radio.
Her honors include a 2025 Gold Stevie Award for Most Innovative Woman of the Year in the health care category and recognition on Becker’s Women in Health IT to Know from 2023 to 2026. A Nashville resident, she won a bronze medal in Olympic weightlifting at the 2015 Masters World Cup. She shares updates on LinkedIn.
