The United States spent an estimated $5.6 trillion on healthcare in 2025, more than any other country on earth. That figure dwarfs peer nations on a per-capita basis. Yet Americans have shorter life expectancies, higher rates of preventable disease, and worse chronic condition outcomes than citizens in most comparable countries.
The gap between spending and results is not a mystery. Research from the Journal of the American Medical Association (JAMA) and the Centers for Medicare and Medicaid Services (CMS) establishes that up to 30% of all U.S. healthcare spending qualifies as waste, roughly $1.6 trillion annually as of 2025. If U.S. healthcare waste were its own economy, it would rank as the 15th largest in the world, exceeding the GDP of Spain, Indonesia, and Switzerland.
For the organizations that fund much of this care, self-insured employers providing health benefits to more than 160 million Americans, the waste is neither abstract nor inevitable. It is measurable, identifiable, and increasingly addressable through artificial intelligence applied at scale across claims data, pharmacy spend, and utilization patterns.
This article examines where the waste originates, why traditional oversight has failed to contain it, and what AI-driven tools are doing right now to close the gap.
The Anatomy of $1.6 Trillion in Annual Waste
A landmark 2019 JAMA study led by Dr. William Shrank identified six major waste categories in U.S. healthcare, estimating total annual waste at $760 billion to $935 billion at that time. Adjusted for CMS 2025 expenditure data, those figures now project to approximately $1.6 trillion per year.
Billing errors and fraud account for the largest share at roughly 25%, or $400 billion annually. The National Health Care Anti-Fraud Association (NHCAA) estimates that between 3% and 10% of all healthcare spending is lost to fraud each year. Industry analyses suggest up to 80% of hospital bills contain errors. The American Medical Association has documented a 20% claims-processing error rate among commercial health insurers, representing an estimated $17 billion in annual waste.
Administrative complexity represents another 22% of waste, approximately $352 billion annually. The JAMA study attributed $265.6 billion to this category alone. The U.S. spends over $1,000 per capita on healthcare administration, roughly five times what comparable nations spend. Unnecessary services, pricing failures, poor care coordination, and clinical inefficiencies account for the remaining waste.
Dr. Shrank’s conclusion was direct: proven interventions could save $191 billion to $282 billion annually. The question has never been whether savings are possible. The question is whether the organizations paying these bills have the tools and the will to find them.
Why Manual Oversight Cannot Solve The Problem at Scale
The traditional response to claims waste has been periodic manual auditing. A third-party auditor reviews a random sample of claims, typically once per year, and reports findings to the plan sponsor. In a health plan processing hundreds of thousands of claims annually, sampling 5% to 10% of transactions leaves the vast majority of claims unexamined.
Conflicts of interest compound the problem. Many third-party administrators (TPAs) process claims on behalf of employers while also being compensated by the same vendors whose charges they are reviewing. Pharmacy benefit managers (PBMs) operating on opaque pricing models profit when drug prices rise, not when they fall. Under these arrangements, waste no longer represents a failure of the system. It represents the system functioning as designed.
Former CMS Administrator Dr. Donald Berwick estimated that as much as $800 billion in waste sits untapped across the healthcare system as a reservoir for potential savings. The money is not hidden. It is sitting in claims data that most plan sponsors never fully analyze, in contracts that most employers never really understand, and in pharmacy arrangements that most organizations accept without scrutiny.
AI changes the equation of oversight. Where a human auditor can review thousands of claims in a week, a machine learning system can review millions in hours. Where a random sample catches a fraction of errors, an AI system processing 100% of claims in near real time catches all errors when programmed correctly.
What AI Is Doing Right Now
The most immediate and measurable application of AI in healthcare waste reduction is in claims auditing. AI-powered systems can review every line item on every claim, compare billed amounts against regional benchmarks, Medicare rates, and contract terms, and flag duplicate charges, unbundled services, upcoding, and charges for services not rendered, all in near real time.
The scale of what these systems can detect is significant. Cotiviti, a major healthcare analytics firm, identified approximately $15 billion in suspect claims across its fraud, waste, and abuse client base in 2023 alone. Matthew Hawley, EVP of Payment Integrity Operations at Cotiviti, stated at the 2024 National Health Care Anti-Fraud Association conference that “fraud, waste, and abuse remain a major challenge to healthcare payment integrity, costing payers and the entire healthcare system billions of dollars each year.” He continued that “as these schemes continue to persist and evolve, it’s critical for plans to adopt artificial intelligence and other advanced technologies that can identify and mitigate FWA while preventing further losses.”
The Centers for Medicare and Medicaid Services confirmed their own deployment of AI tools in a 2023 statement: “We use artificial intelligence and machine learning to find potential fraud that would not be apparent to the human eye. We try to use the latest technology to make potential fraud easier to detect more quickly.” CMS addresses confirmed schemes through vulnerability analyses, law enforcement referrals, and regulatory action.
Predictive modeling represents a second high-impact application. Predictive modeling engines analyze historical claims data, demographic information, clinical indicators, and pharmacy utilization to predict which plan members are likely to become high-cost members in the coming months, which frequently indicates those more likely to need more care in the immediate future. This allows care managers to intervene before a chronically ill member’s condition becomes a $200,000 hospitalization. Prevention at this scale is not achievable through manual data review.
Price transparency data, now powered by payer machine-readable files mandated by federal legislation, give employers the ability to build market-by-market benchmarks for every billing service or procedure code. AI systems can transform this raw data into powerful, actionable comparisons, identifying where a plan is overpaying relative to the market and steering members toward high-value providers. David Pierre, CEO of a major payment integrity entity formed from the merger of three specialized analytics firms, described the technology’s potential in a 2024 interview: “Now you have technology, machine learning, AI that can look at things, serve it up to the experts so they can actually make the decision… I think this will really change the trajectory of healthcare spend across the U.S.”
The Fiduciary Dimension
AI adoption in healthcare is accelerating in part because the legal environment for passive plan management is changing. Under the Employee Retirement Income Security Act (ERISA) and the Consolidated Appropriations Act (CAA) of 2021, self-insured employers have a legal fiduciary duty to ensure plan assets are spent prudently and in the best interest of plan participants.
That obligation is no longer theoretical. In 2024 and 2025, Johnson & Johnson, Wells Fargo, JPMorgan Chase, and Mayo Clinic each faced lawsuits alleging fiduciary breaches related to prescription drug pricing, excessive fees, and inadequate vendor oversight. A National Alliance for Healthcare Purchaser Coalitions survey found that 65% of employers expressed growing concern about potential litigation related to fiduciary duties.
AI-powered analytics tools are being positioned by compliance-conscious plan sponsors as evidence of prudent oversight. A plan sponsor who can demonstrate that 100% of claims are being reviewed, that vendor contracts have been sanitized of opaque, problematic clauses, that provider pricing is benchmarked against market data, and that pharmacy utilization is audited on an ongoing basis is in a materially different legal position than one relying on annual, random sample reviews.
Elizabeth Mitchell, President and CEO of the Purchaser Business Group on Health (PBGH), articulated the employer obligation directly: “Under the Consolidated Appropriations Act of 2021, employers are legally accountable as fiduciaries for their health plans, requiring them to provide employees with the best healthcare benefits for the best price.” Technology is now the clearest path to meeting that standard.
Where AI in Healthcare Is Headed
The current generation of AI applications in healthcare focuses primarily on detection: identifying billing errors, fraud patterns, utilization anomalies, and pricing outliers, largely after the fact. The next generation is shifting toward prediction and prevention.
AI systems can be trained on multi-year claims datasets to model chronic disease trajectories down to the individual member level. When a predictive model flags a member as high-risk for a cardiovascular event six months before it would otherwise appear in claims data, a care manager can intervene proactively to provide needed help. It’s important to note that the model itself does not make clinical decisions; it merely surfaces risk factors that enable humans to act.
Natural language processing is being applied to prior authorization workflows, reducing the 13 hours of weekly staff time that physician practices currently spend on authorization requests. Automated processing of routine authorizations, combined with AI pattern recognition for appropriate clinical criteria, can accelerate approval timelines and reduce the 93% physician-reported rate of care delays attributed to prior authorization.
Pharmacy analytics are becoming more precise. AI systems can identify members taking branded medications where clinically equivalent, lower cost generics are available. They can flag generic therapeutic alternatives that cost a fraction of brand-name generic drugs, and can detect formulary patterns suggesting PBM incentive misalignment rather than clinical optimization. Organizations transitioning to transparent, pass-through pharmacy pricing models report savings of 15% to 30% on their pharmacy spend, according to research from Model Optimal Care.
From Detection to Decision
In an ideal scenario, AI systems generate findings and humans act on them. The most effective deployments of AI in healthcare waste reduction pair analytical precision with human-driven decision making. Under this scenario, claims audit results lead to contract renegotiation, predictive modeling flags activate care management outreach, and pharmacy analytics drive formulary redesign.
Bill Fera, Principal Consultant at Deloitte, captured the dynamic clearly: “We’re taking the mystery away. And there is a fact base. There’s a core piece of information that can be interrogated. It’s just now, it can be interrogated very quickly.” Speed and complete coverage are what AI brings to the table. A claims dataset that would take a human team months to analyze can be processed by AI overnight, with findings ranked by financial impact and presented for decision-making the following morning.
For self-insured organizations funding health coverage for over 160 million Americans, this is not a technology conversation in isolation. It is a financial management conversation. The employer that deploys AI-powered claims auditing on a $50 million health plan and recovers funds on a 14% payment inaccuracy rate captures over $7 million annually that would otherwise fund vendor profit rather than employee benefits.
The $1.6 trillion in annual U.S. healthcare waste is not distributed evenly across the system. It is concentrated in identifiable patterns: duplicate charges, inflated facility fees, upcoded and unbundled procedures, spread pricing by PBMs, and avoidable or redundant utilization. AI will not eliminate the incentive structures that produce waste. What it can do is make the sources of waste visible, measurable, and actionable at a scale that fundamentally changes what plan sponsors can demand from their vendors and their data. That is a step in the right direction.
About the Author
Jude Odu is the Founder of Health Cost IQ, a healthcare analytics company that serves self-insured organizations. He is the author of Model Optimal Care: End U.S. Healthcare Waste, One Health Plan at a Time, scheduled for release in April 2026. He has over 25 years of experience in healthcare technology across payers, providers, software, and academia. Learn more at judeodu.com.
