UpEvidence

Where AI Can Take Cost Out of Health Care

A recent analysis estimates where artificial intelligence could reduce annual healthcare spending, with results separated by stakeholder, operating domain, and administrative versus medical expense.

Home / UpEvidence / Where AI Can Take Cost Out of Health Care

Explore the estimates

Across U.S. health care, the authors estimate that AI could reduce annual spending by 5.7% to 10.6%, or approximately $439 billion to $811 billion. Use the controls below to examine where the estimated opportunity sits. Percentages represent the share of annual spending in each operating domain.

Hospitals 8.6% to 14.4% $131B to $220B annually

Physician practices 2.8% to 8.6% $29B to $90B annually

Private insurers 7.0% to 12.1% $205B to $357B annually

All U.S. health care 5.7% to 10.6% $439B to $811B annually

Stakeholder

Search domains and AI use cases

Administrative expense Medical expense Shared scale: 0% to 16%

What the estimates do and do not show

The analysis sizes direct savings within each stakeholder group. It does not by itself show how a change for one stakeholder affects another. For example, reduced medical spending for an insurer may change provider revenue, patient access, utilization patterns, staffing needs, or contract incentives. Those second-order effects require additional data, assumptions, and modeling.

Our perspective

We appreciate that the authors quantify the first-level effects by stakeholder group, domain, and expense type (administrative and medical). If anyone wants to analyze second-order effects, reach out to FastHSR.

Source

Sahni NR, O'Gorman K, Agarwal R, Bello Thornhill H, Cutler D. The Value Opportunity from Artificial Intelligence in U.S. Health Care Spending. NEJM Catalyst. 2026;7(9). doi:10.1056/CAT.26.0037.

More UpEvidence