AI Scribes Were Supposed to Save Healthcare Money. They Added $2.3 Billion in Costs Instead.

Hospitals spent $2.3 billion more on healthcare claims over three years because of AI scribes. Not despite AI scribes. Because of them.

Blue Health Intelligence, the data analytics arm of the Blue Cross Blue Shield Association, analyzed commercial claims data from Q2 2022 through Q1 2025. The finding: hospitals using AI-powered coding and documentation tools generated $663 million in additional inpatient spending and at least $1.67 billion in additional outpatient spending. That is not a rounding error. That is a systemic cost shift paid by insurers, employers, and ultimately by you through higher premiums.

Meanwhile, a separate analysis from Trilliant Health examined six large health systems across 13 states that publicly adopted AI scribes between 2018 and 2024. At one health system, high-intensity new patient visit coding jumped from 60.5% to 80.0% of all visits. The shift ranged from 12 to 20 percentage points across all six systems.

The myth that AI saves money in healthcare just ran into $2.3 billion worth of evidence.

What AI Scribes Actually Do (And Why It Matters for Your Insurance Bill)

AI scribes are ambient listening tools that sit in the exam room during doctor-patient conversations. They record everything, generate clinical notes, and suggest billing codes. Products like Microsoft's DAX Copilot, Abridge, and Nuance dominate the market. Microsoft/Nuance holds 33% market share. Abridge holds 30%. Together, they cover most major hospital systems in the United States.

The pitch is simple: doctors spend 2 hours on paperwork for every 1 hour with patients. AI scribes handle the documentation. Doctors get their evenings back. Provider burnout drops from 51.9% to 38.8% within 30 days of adoption.

That part is real. Burnout reduction is measurable and significant.

But here is what nobody put in the marketing brochure: when you record every word a patient says and feed it through an AI trained to maximize documentation completeness, the resulting clinical notes capture more diagnoses, more complexity, and more billable detail than a tired doctor typing notes at 9 PM ever would. More documentation means higher-intensity billing codes. Higher-intensity codes mean bigger insurance claims.

The tool sold as a burnout reducer is functioning as a revenue maximizer.

The Numbers: How Coding Intensity Shifted at Six Major Health Systems

Trilliant Health's analysis tracked evaluation and management (E/M) billing codes at six anonymized health systems (labeled A through F) from 2018 to 2024. The results show a consistent pattern across every single system.

New Patient Visits: 12 to 20 Percentage Point Shifts

High-intensity new patient codes (CPT 99204-99205) increased at every system studied:

• Health System C: 60.5% to 80.0% (+19.5 points) — the most dramatic shift
• Health System E: 47.5% to 67.0% (+19.5 points)
• Health System D: 44.5% to 63.7% (+19.2 points)
• Health System F: 43.7% to 60.2% (+16.5 points)
• Health System B: 42.9% to 57.6% (+14.7 points)
• Health System A: 51.3% to 64.8% (+13.5 points)

Before AI scribes, roughly 40-60% of new patient visits were coded as high-intensity. After adoption, that number jumped to 57-80%. Health System C now codes 4 out of every 5 new patient visits at the highest billing level.

Established Patient Visits: 7 to 12 Percentage Point Shifts

The pattern repeats for return visits (CPT 99214-99215), though the magnitude is smaller:

• Health System B: 40.9% to 52.7% (+11.8 points)
• Health System F: 47.8% to 57.7% (+9.9 points)
• Health System D: 50.4% to 60.2% (+9.8 points)
• Health System E: 56.6% to 64.6% (+8.0 points)
• Health System C: 65.3% to 72.9% (+7.6 points)
• Health System A: 59.8% to 67.2% (+7.4 points)

Every system shifted upward. Not one went down. Not one stayed flat. The direction is unanimous.

Specific Diagnosis Coding Spikes

The Blue Health Intelligence analysis dug into specific diagnoses. Among hospitals with the fastest AI adoption, acute posthemorrhagic anemia coding in maternity admissions rose from 4% to 12.3%. At hospitals with slower AI adoption, the same diagnosis only increased from 7.9% to 8.2%.

Here is the critical detail: transfusion rates at the high-growth hospitals barely moved (0.7-0.9% increase). If more patients actually had severe anemia requiring treatment, you would expect more blood transfusions. The transfusions did not follow the diagnoses. The diagnoses increased without a corresponding increase in treatment.

Dr. Razia Hashmi, BCBSA's Vice President of Clinical Affairs, was direct: "Something is disconnected. Among hospitals showing the fastest rise in diagnoses of post-partum anemia, the rise in patients coded with this condition wasn't paired with the level of care we would have expected, and the patterns we're seeing point to AI-enabled coding."

The $600 Million Market Nobody Talks About Honestly

The AI ambient clinical documentation market was valued at $600 million in 2025. It is projected to reach $27.8 billion by 2034, a compound annual growth rate of 48.2%. Healthcare AI adoption in the U.S. jumped from 3% to 22% in just two years. At UCSF, 70% of physicians now use AI scribes daily.

These tools cost hospitals $99 to $299 per provider per month, with the American Academy of Family Physicians estimating an average of $150 to $200 per provider per month. Enterprise solutions like DAX Copilot run higher. For a 500-physician hospital system, that is $75,000 to $150,000 per month in AI scribe licensing fees.

But the revenue those tools generate through higher-intensity coding dwarfs the subscription cost. The top 10% of hospitals by coding growth showed 13.1 percentage points higher complex-coded admissions compared to 4.2 percentage points at the remaining 90%. Those top hospitals reached 59.8% complex-coded admissions by Q1 2025.

Do the math: a hospital pays $1.8 million per year for AI scribes across 1,000 providers. The same hospital generates millions more in higher-coded claims. The ROI is not in saving doctors time. The ROI is in billing optimization that no one explicitly asked for.

Upcoding vs. Improved Documentation: The Convenient Ambiguity

This is where the story gets uncomfortable for everyone involved.

Hospital systems argue that AI scribes capture clinical details that doctors previously missed or did not bother to document. Under this framing, the coding increases represent accurate billing for work that was always being done but never properly recorded. Doctors were systematically undercoding for years, and AI scribes corrected a documentation gap.

Allison Oakes, Chief Research Officer at Trilliant Health, acknowledged this possibility: "These AI-enabled scribing tools are allowing clinical documentation to be captured more thoroughly and accurately. Because we see this systematic increase in coding intensity, it probably suggests that, historically, providers have actually been under-coding their visits."

Insurers see it differently. When a diagnosis code appears without a corresponding treatment, the coding looks like revenue optimization, not documentation improvement. The postpartum anemia example is their strongest evidence: diagnosis rates tripled at AI-heavy hospitals while transfusion rates barely moved.

The truth is almost certainly both. AI scribes do capture legitimate clinical details that rushed doctors miss. AI scribes also surface every billable condition that a conservative doctor might have left undocumented. The first outcome improves care documentation. The second outcome increases costs without improving care. And there is no clean way to separate them.

Caroline Pearson, Executive Director of the Peterson Health Technology Institute, summarized the private consensus: "The investors, the health plans, and the providers, in private, were like, 'OK, well, it's quite clear scribes are increasing coding intensity. One hundred percent.'"

Everyone knows. Nobody agrees on what to do about it.

What This Means for Your Health Insurance Premiums

Higher-coded claims mean higher reimbursements from insurance companies. Insurance companies pass those costs to employers through higher premiums. Employers pass them to you through higher deductibles, copays, and paycheck deductions.

The $2.3 billion Blue Health Intelligence identified is from commercial claims only — employer-sponsored and individual market insurance. It does not include Medicare or Medicaid, where similar coding patterns likely exist but have not been publicly quantified at this scale.

If the same coding intensity shifts apply to Medicare (which covers 65 million Americans) and Medicaid (which covers 90 million), the total cost impact could be multiples of $2.3 billion. We do not have that data yet, but the Trilliant Health analysis used all-payer claims data showing the same patterns across payer types.

For comparison, the entire U.S. spent approximately $4.5 trillion on healthcare in 2024. A $2.3 billion increase from AI coding tools alone represents a small percentage of total spending. But that number covers only three years of data from commercial claims at hospitals that adopted early. As AI scribe adoption expands from 22% to a projected majority of providers, the spending increase will scale proportionally.

The Regulatory Blind Spot

Federal regulators have been slow to address AI-driven coding intensity. The False Claims Act prohibits billing for services not rendered, but AI scribes operate in a gray zone. The documentation supports the code. The question is whether the documentation reflects genuine clinical complexity or AI-optimized detail capture that inflates perceived complexity.

CMS (Centers for Medicare and Medicaid Services) has not issued guidance specifically addressing AI scribe-driven coding patterns. The OIG (Office of Inspector General) has flagged upcoding concerns generally but has not published audit findings specific to AI ambient documentation tools.

The regulatory gap creates a first-mover advantage for hospitals that adopt aggressively. Early adopters capture higher reimbursements before rules tighten. Late adopters face pressure to match their competitors' coding intensity. The result is an arms race where every hospital needs AI scribes just to keep up with the billing levels their competitors have normalized.

This is how a burnout reduction tool becomes a mandatory billing infrastructure investment. Not because hospitals want to upcode. Because the hospital across town already adopted, and your coding intensity now looks suspiciously low by comparison.

Who Wins and Who Loses

Winners:

• AI scribe vendors — $600M market growing at 48.2% CAGR, projected to hit $27.8B by 2034
• Hospitals — Higher reimbursements per visit with minimal additional clinical effort
• Doctors — Less burnout, more documentation accuracy, and their employers earn more per visit

Losers:

• Insurance companies — Paying higher claims without evidence of improved patient outcomes
• Employers — Absorbing premium increases driven by coding intensity shifts
• Patients — Higher premiums, deductibles, and copays without receiving additional care

The most cynical reading: AI scribes transfer wealth from patients and employers to hospitals and AI vendors. The documentation improvement is real but incidental to the financial outcome.

What Happens Next

Three forces are converging:

Insurer pushback. Blue Cross Blue Shield published this research for a reason. Expect payers to demand AI scribe audit trails, pre-authorization for AI-suggested codes, or flat reimbursement caps that neutralize coding intensity gains.

Regulatory attention. The BCBSA data gives CMS and OIG the evidence base for targeted audits. If post-partum anemia coding tripled without matching transfusion rates, that specific pattern becomes an audit trigger.

Market saturation. When every hospital uses AI scribes, the coding intensity advantage disappears. All hospitals code at the same high level, insurers reprice their plans accordingly, and the net effect is a permanent cost increase baked into baseline healthcare spending.

The irony is structural. AI scribes genuinely reduce physician burnout. They genuinely improve documentation completeness. And they genuinely increase healthcare costs. All three things are true simultaneously. The question is whether the burnout reduction justifies the cost increase — and who should pay the difference.

For a broader look at how AI tools are reshaping professional workflows beyond healthcare, check out the AI tools directory on Skila. If you are interested in the open-source infrastructure behind ambient AI systems, explore AI repos and MCP servers that power similar documentation automation. And for more analysis of AI's economic impact across industries, follow Skila AI News.

Frequently Asked Questions

What are AI scribes in healthcare?

AI scribes are ambient listening tools that record doctor-patient conversations during clinical visits and automatically generate clinical notes and billing codes. Major vendors include Microsoft's DAX Copilot (33% market share), Abridge (30%), and Ambience Healthcare (13%). They cost $99-$299 per provider per month and are now used by 22% of U.S. healthcare providers.

How much have AI scribes increased healthcare costs?

Blue Health Intelligence found $2.3 billion in additional healthcare spending from AI-enabled coding tools between Q2 2022 and Q1 2025, based on commercial claims data. This breaks down to $663 million in inpatient spending and at least $1.67 billion in outpatient spending. The actual total is likely higher when Medicare and Medicaid claims are included.

Is AI scribe upcoding illegal?

Not necessarily. The False Claims Act prohibits billing for services not rendered, but AI scribes generate documentation that technically supports higher billing codes. The debate centers on whether the AI-captured documentation reflects genuine clinical complexity or inflates perceived complexity through exhaustive detail capture. Federal regulators have not issued specific guidance on AI scribe-driven coding patterns as of April 2026.

Do AI scribes actually reduce physician burnout?

Yes. Provider burnout decreased from 51.9% to 38.8% within 30 days of AI scribe implementation, according to research cited by Trilliant Health. AI scribes eliminate 1-2 hours of daily documentation work per physician. The burnout reduction is real and well-documented — it coexists with the cost increase rather than contradicting it.

What are the best alternatives to AI scribes for clinical documentation?

Traditional alternatives include human medical scribes ($36,000-$45,000 per year per scribe), voice-to-text dictation tools like Dragon Medical, and templated EHR documentation. However, none match AI scribes for hands-free ambient documentation. The real question is not whether to use AI scribes but how to prevent AI-driven documentation from systematically inflating billing codes without corresponding clinical value.