AI is reshaping healthcare economics, just not the way the demos promise

Every healthcare pitch deck now has an AI slide (if AI hasn't already been added to the name or company tagline). Most describe a future in which software reads the chart, codes the claim, engages with the patient and books the visit, while margins expand and headcount falls.

The reality in 2026 is narrower and more useful. AI is already changing healthcare services, but not mainly by replacing clinicians. It is changing who gets paid, how fast and how much. For investors and operators, that distinction matters more than any slick demo.

Where it works today: Documentation

The clearest win so far is ambient documentation: AI that listens to a visit and drafts the note. It targets a real problem. Physicians spend hours a day on documentation, much of it after the last patient leaves.

The best evidence is encouraging but modest. A UCSF study of nearly 1.2 million ambulatory encounters across 1,565 physicians found that AI scribe users generated 5.8% more work RVUs and saw 2.8% more patients per week than non-users. At 2025 Medicare rates, that was worth about $3,044 in additional revenue per physician per year. Claim denial rates did not change.

Those are real gains in measured productivity. But tyhey are not a transformation. The study covered early adopters at a single health system, and its authors noted that some of the lift may reflect better coding rather than more care. For an operator, AI documentation is a retention tool and a modest productivity lever. It is not a reason to pay a higher multiple.

Where it matters most: The Revenue Cycle

The larger shift is in the revenue cycle, and it is not happening on the provider’s schedule.

Payers are using AI to review claims and prior authorization requests at a scale no human team could match. Insurers denied more claims on clinical grounds in 2025 than in 2024, contributing to a 25% increase in net revenue leakage at hospitals, according to Healthcare Finance News. Payers now review full claim datasets rather than samples, and one national insurer processes 80% of prior authorizations in real time.

Traditional Medicare has joined in. The CMS WISeR model, which began January 1, 2026 and runs through 2031, uses technology vendors that combine AI with clinician review to handle prior authorization for selected services in six states: Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. The vendors are paid a share of the spending they help avert.

The result is an arms race. Payers deploy AI to find reasons not to pay; providers deploy AI to document, code and appeal. Large health systems can afford both sides of the fight. Smaller providers—the middle-market businesses private equity buys—often cannot. For them, AI-enabled revenue cycle and payment integrity services are no longer optional. They are how a provider keeps what it has already earned.

What AI will not do

AI will not solve healthcare’s workforce problem. It can make a therapist, nurse or physician somewhat more productive. It cannot create one. Psychiatrist supply is projected to fall 12% by 2037 while demand for mental health care rises 44%, and no software closes that gap. Most healthcare services are still delivered in person, by licensed people, under rules written by and for them.

AI also will not protect a business whose moat is software alone. In the first half of 2026, what Bain calls the AI-driven “SaaSpocalypse” muted healthcare IT deal activity, as investors questioned what software features are worth when AI can replicate them. The durable advantage is shifting to businesses that pair technology with proprietary data, embedded workflows and accountability for outcomes.

What investors should ask

For sponsors, the useful question is not whether a company “uses AI” or how. It is where AI changes the unit economics, and who captures the benefit. In diligence, that means asking:

• Where does the business lose revenue today—denials, underpayments, missed charges—and what is it doing about it?
• How much clinician time goes to documentation and administration, and has the company measured what automation returns?
• Is the business on the right side of the payer arms race, or counting on payers not to use the tools they already have?
• Does the company own data or workflows that improve with scale, or is it reselling a feature someone else will soon give away?

The Brutal Take

AI is not a thesis. It is a tool, one that makes good operators better and exposes weak ones faster. Its biggest near-term effect on healthcare services will be financial, not clinical: it is rewriting the rules of who gets paid and how quickly.

The businesses that win will use it the way the best operators use any tool: deliberately, measured against results, and in service of the patients and clinicians the business exists for. The ones that treat it as a slide in the deck will find their payers took it more seriously than they did.

‍