For years, AI has been promised as a “silver bullet” that will solve exploding healthcare costs, yet new data reveals it’s doing the exact opposite. While public attention focuses on “AI doctors,” a staggering 89% of healthcare’s total AI spend is being funneled into back-office tools designed to maximize provider revenue.
Our data suggests that these tools are responsible for a 1.7% annual increase in total employer healthcare spend, out of an estimated 9% total increase in 2026. To combat this, it is more important than ever that benefits leaders steer their employees to high-value providers who consistently deliver excellent quality care at the right price.
ACCESS THE FULL WEBINAR AND REPORT: 'How Much is AI Upcoding Costing Employers?' to dive deeper.
Figure 1: The primary drivers of annual healthcare cost trends broken down by category.
How is Healthcare Using AI?
The healthcare sector has been the biggest adopter of AI in the US, deploying at a rate 2.2x faster than the broader economy. In just the last year, healthcare’s total spend on AI services and licenses tripled, reaching $1.4B.
This investment is broken into 3 main categories: (1) front office tools, (2) AI care providers, and (3) back office tools.
While public attention focuses on patient engagement - including both front office tools and AI care providers, the majority of capital is being deployed for back office tools in areas like administrative efficiency and revenue cycle management.
A staggering 93% of healthcare’s total spend on AI licenses and services last year was directed towards back office tools – with almost all (89%) going to provider tools, and only a small subset (4%) payer tools. The remaining 7% went towards all of patient engagement (including front office tools and AI care providers) combined.
Figure 2: Breakdown of healthcare industry spending on healthcare-specific AI services and licenses by category in 2025. These figures do not include any VC or external investments directed towards startups.
The AI Arms Race: Providers vs. Payers
All of the funding flowing into back office tools is creating a technological ‘arms race’ where providers are using AI to claim every possible dollar and payers are using their own AI to find reasons to deny more claims. Claims denials have steadily increased year over year, with reported denial rates up 13% over the past three years. Payers specifically tie these rising denial rates - enabled by AI - to an increase in upcoding from providers. 80% of payers expect a continued increase in both codes per claim and total claim volume as a result of provider AI tools.
How AI Upcoding Appears in Garner’s Data
This arms race is not a theoretical risk – it’s already appearing in employer claims data. And unsurprisingly, given the imbalance in investment dollars between providers and payers, providers are clearly winning. Garner’s analysis found that AI-enabled billing tools are responsible for a 1.7% annual increase in total employer healthcare spend in the last year.
AI upcoding is impossible to detect in individual cases or small sample sizes. The only way to see AI upcoding in the data is by looking at large scale aggregated datasets. And when we do, we see small shifts in coding behavior, applied across millions of claims, that are meaningfully increasing employer healthcare spend.
We first see this in systemic “severity inflation”. Because AI can scan a doctor’s notes and suggest higher-complexity codes that a human might have missed, we are seeing a steady creep in billed intensity across the board.
Figure 3: Average coded level of office visits over time (LHS) and ER visits (RHS) and their impact on healthcare cost.
From 2021 to 2025, the average coding level of office visits increased by approximately 5%, leading to a .57% increase in total healthcare costs. ER coding levels also increased by approximately 5%, leading to a .28% increase in total healthcare costs. These increases reflect higher billed complexity, not necessarily more complex care.
In addition to ‘severity inflation’, AI tools enable many additional billing strategies that result in an artificial increase in ‘utilization’ of services. For example, we see a meaningful increase in the use of Modifier 59, a claims modifier that allows procedures to be billed separately that would otherwise have been bundled. In addition, we see the average number of codes per claim increasing as AI tools identify additional billing opportunities. Crucially, we do not believe either of these billing trends actually mean more care is being delivered – this is the same care being delivered, with new AI-enabled billing tactics being used to drive up revenue.
Figure 4: Percent of surgeries with an attached Modifier 59 over time (LHS) and the average number of diagnoses codes per claim over time (RHS) and their impact on healthcare trends.
These examples are just a few of the ways AI tools can lead to higher healthcare costs, even when more medical care is not being provided. Other examples include Diagnosis-Related Group intensity upcoding and automated appeals, which also lead to higher claim volume and greater costs. These factors, coupled with severity inflation and greater perceived increase in utilization, and others factors, lead to a 1.7% increase in annual employer healthcare spend.
The Real Opportunity: Provider-Level Performance
At Garner, we believe the key to controlling healthcare costs lies in guiding patients to the best doctors. And this holds true even in the face of AI upcoding.
Not all providers use AI tools in the same way. Claims data shows significant variation in quality, efficiency, and coding behavior at the individual provider level.
Figure 5: Breakdown of performance by the top and bottom 25% of providers across intensive site of service, inappropriate prescriptions, post-surgery complications, and upcoded office visits.
Guiding patients towards doctors who are billing appropriately – doctors who can deliver excellent quality care without the upcoding - is the key to tackling this problem. Doing so not only avoids cost in the near-term, but creates pressure on providers to avoid AI upcoding, lest they risk losing significant patient volume. The result is a virtuous circle where patients see doctors who can deliver the best care, at the right cost, and doctors know that if they deliver the best care, at the right cost, they will be rewarded with more patient volume.
Future Implications for Employers
The transition to AI-driven healthcare documentation is more than an administrative change—it is a fundamental shift in how provider revenue is generated. Failing to address "provider upcoding" as a core driver of spend risks leaving your health plan vulnerable to automated inflation, especially as adoption of these tools, which is relatively modest today, scales rapidly in the coming years.
Benefits leaders can no longer just "manage the carrier," and instead must proactively take steps to help protect their plan from algorithmic documentation.
By identifying the high-value providers who prioritize clinical evidence over billing optimization—and giving employees a clear, incentivized path to find them—employers can turn the tide on healthcare's 9% annual trend.
As healthcare costs continue their upward trajectory, the ability to identify high-value providers who practice efficient, evidence-based medicine will increasingly dictate an employer's ability to control costs without sacrificing care quality.
ACCESS THE FULL WEBINAR AND REPORT: 'How Much is AI Upcoding Costing Employers?' to dive deeper.
Figure Sources:
Figures and calculations derived from Garner Health proprietary analytics of the nation’s largest claims dataset integrated with third-party datasets from the U.S. Federal Reserve, Harvard University, and the Kaiser Family Foundation. Additional industry benchmarks provided by the Business Group on Health, Menlo Ventures, the American Journal of Medicine, and national insurance carriers.