Experts Weigh Cost-Saving Promises of AI in Health Care Against Risk of Higher Spending

Artificial intelligence (AI) is widely touted as a solution to health care’s cost and complexity problems, but whether it will actually bend the spending curve remains an open question.

That tension—between AI’s promise and its potential to exacerbate waste—anchored a discussion at the 2026 virtual V-BID Summit during a session on “Employing Innovative Technology to Enhance Value and Reduce Waste.”

Framing AI Within an Affordability Crisis

The panel featured Michael Chernew, PhD, director of the Healthcare Markets and Regulation Lab, Harvard University; Hannah Neprash, PhD, a health economist and associate professor, University of Minnesota; and Jeremy Friese, MD, MBA, founder and CEO of Humata Health. The discussion focused on AI’s influence on administrative efficiency, clinical care, and prior authorization, with speakers emphasizing that technology alone will not determine whether AI improves value in the health care system.

Chernew, who is also the co–editor in chief of The American Journal of Managed Care^®, opened the session by placing the conversation within what he described as a growing affordability crisis. National health expenditures have been increasing by approximately 7% annually and are projected to grow between 5% and 6% in the coming years, a trend he said is unsustainable without meaningful changes to the system.

The debate comes amid longstanding concerns about administrative inefficiency in the US health system. Estimates suggest administrative activities—including billing, insurance-related tasks, and regulatory compliance—account for roughly 15% to 25% of total US health care spending, representing hundreds of billions of dollars annually.¹ These costs are significantly higher than those in peer nations and are frequently cited as a target for automation and digital tools such as AI.

Although expanding access to high-value services remains a priority, Chernew cautioned that the concept of value cannot distract from broader fiscal realities.

Four Domains Where AI Is Reshaping Care

Chernew outlined 4 areas where AI is already influencing health care delivery and spending: administrative processes, physician support, patient support, and drug development.

Administrative functions represent one of the most immediate applications. AI tools can automate tasks in human resources, customer service, and claims review, while also helping payers identify unusually aggressive coding patterns. However, similar technology can also assist providers in documenting additional comorbidities and selecting higher reimbursement codes, making the overall financial impact uncertain.

AI may also reshape clinical workflows. Algorithms can help triage imaging studies, flag urgent findings, and reduce certain low-value services. In addition, patient-facing AI tools could guide some individuals toward lower-acuity care settings, potentially reducing unnecessary emergency department visits.

Yet these efficiencies could have countervailing effects. By making imaging interpretation faster and more accessible, AI could increase overall imaging volumes, detect additional abnormalities, and lead to more follow-up procedures and interventions.

Drug development represents another area of potential transformation. AI-assisted discovery could accelerate the creation of therapies that prevent expensive downstream complications. At the same time, these tools may enable increasingly personalized—and potentially costly—treatments that add to overall spending even if patient outcomes improve.

“I believe strongly that the health care system needs to provide access to high-value services to patients that need them, but I think that we can't use the appeal to value as an excuse to ignore the core fiscal challenges,” said Chernew.

Administrative Relief, Diagnostic Gains—and Care Cascades

Neprash focused on the broader system-level implications of AI adoption. She emphasized that there is still limited evidence on the technology’s overall effect on health care budgets.

“If anybody tells you that this is an obvious question with an obvious answer, they're lying to you,” said Neprash. “I really do think that we don't know yet what AI will do to system-level health care spending, but I want to start with the glass-half-full story here and share a few ways and a few specific applications in which AI has the potential to either reduce health care spending or at least slow down the growth of spending.”

Neprash pointed to AI-enabled automation in call centers, prior authorization systems, and clinical documentation. Generative AI scribes that listen to clinical encounters and draft notes have already improved clinician satisfaction in some settings, potentially reducing burnout and costly workforce turnover.

Some analysts believe AI could significantly reduce waste if applied to administrative processes and clinical workflows. Estimates from industry analyses suggest AI and related technologies could save between $200 billion and $360 billion annually in US health care spending, largely through improvements in automation, clinical decision support, and operational efficiency.² However, experts caution that realizing these savings will depend heavily on how the technology is integrated into payment models and care delivery.

AI can also support diagnostic decision-making. For example, algorithms that analyze CT images to produce 3D models of coronary arteries can help identify areas of stenosis in patients with suspected coronary artery disease. These tools may allow clinicians to substitute noninvasive imaging for more invasive tests and initiate earlier interventions such as lifestyle modification or statin therapy.

Still, Neprash warned about the risk of care cascades. AI tools that provide a second set of eyes may detect more polyps during colonoscopy or uncover incidental findings when existing CT scans are reanalyzed for unrelated conditions, such as low bone mass. Although these findings can be clinically meaningful, they may also lead to additional testing, procedures, and patient anxiety without clear benefit.

AI Coding Wars and the Promise of Smarter Prior Authorization

Neprash also highlighted what she described as emerging “AI coding wars.” Revenue cycle software increasingly uses AI to optimize billing codes and document disease severity, while payers deploy their own algorithms to identify potential overcoding and apply downcoding strategies.

Friese addressed prior authorization as an area where AI could reduce administrative friction without creating the same adversarial dynamics. As a former academic radiologist, he described health systems relying on “warehouses of humans” to manage prior authorization requests through faxes and web portals.

Humata’s AI platform integrates with electronic medical records to determine whether prior authorization is required and automatically assemble the supporting documentation when it is needed.

The company participates in the Wiser model from the Center for Medicare and Medicaid Innovation, which uses AI to review a limited set of outpatient services in 6 states. Requests that clearly meet national or local coverage determinations can be approved immediately by the system, while more complex cases are referred to human clinicians for review.

“When the provider has AI, it benefits the payer,” said Friese. “When the payer has AI, it actually benefits the provider with one key important point, and that is the AI can never say no. And I'm a firm believer that when on the payer side, AI should never say no. It should take a doc, a nurse, [or] someone that actually can understand the consequences of health care to finally make those calls.”

Across the discussion, panelists agreed that AI’s ultimate impact on health care value will depend less on the technology itself and more on the payment models, benefit designs, and regulatory frameworks that shape how it is used. Without those guardrails, they warned, AI could just as easily amplify existing inefficiencies as reduce them.

References

1. Chernew ME, Mintz H. Administrative expenses in the US health care system: why so high? JAMA. 2021;326(16):1495-1496. doi:10.1001/jama.2021.16461

2. CAQH Index: $13.3 billion—33% of healthcare administrative spending—could be saved annually through automation. CAQH. Published January 21, 2020. Accessed March 12, 2026. https://www.caqh.org/blog/caqh-2019-index-133-billion-33-percent-healthcare-administrative-spend-can-be-saved-annually