Unbiased headlines. Facts, not spin.
Unbiased Headlines is an independently owned news organization. We bring you the facts, not opinions, written fresh from 110+ named sources. Sources linked so you can verify it yourself.
Health Plans Project 9% Cost Rise in 2027 and Name Providers' AI Billing Tools Among the Reasons

The pitch for AI in healthcare has been simple: less paperwork, faster diagnoses, lower costs. The cost numbers heading into 2027 point the other way.
Health plans responding to PwC's annual medical cost trend survey project that commercial healthcare costs will climb 9% in 2027. They name providers' adoption of AI-enabled revenue optimization tools among the leading reasons. Benefits broker Relation Insurance, summarizing the survey, says nearly 70% of health plans rank those tools among their top three cost inflators for 2027.
The mechanism is documentation and coding. Relation Insurance says the tools help providers capture and bill for more of the complexity in each visit, which raises the cost per claim even when the care delivered has not changed. That is the plans' account. Providers have not been quoted disputing it in the material reviewed for this story.
The money going in
The spending is real. Menlo Ventures, a venture capital firm that invests in AI, released a report today, Nov. 16, putting healthcare AI spending at $3.7 billion, up 2.5x in a year. Menlo says that is more than legal and finance combined.
Menlo says Microsoft and Epic now capture roughly two of every three new AI scribe dollars. It also says the fastest growth has moved beyond paperwork: life sciences AI spending rose 23x, clinical copilots 5.5x and patient engagement 4.5x.
"AI is transforming healthcare," said Greg Yap, a Menlo venture partner, adding that the sector is driven by AI's potential "to improve health outcomes and lower costs for everyone." Menlo sells to and invests in this market, so those are the claims of an interested party.
The payoff is harder to find
Bain & Company surveyed health system administrators and found a gap between ambition and results. Nearly two-thirds expect meaningful, quantifiable returns within three years. More than 60% put between 5% and 25% of their IT budgets into AI.
But only 40% say they can clearly measure the return. About 15% report cost savings of 10% to 20%. Most have seen thinner results.
Bain points to three problems:
- Workflows: More than half of administrators say they are adding tools to solve isolated problems. Only 10% have redesigned core workflows end to end with AI built in.
- Data: Only 15% of clinicians say they have access to the data they need to deliver optimal care, and fewer than a third of organizations have a funded roadmap to close the gaps.
- Trust: Clinicians cite doubts about AI's accuracy, unclear accountability when something goes wrong, and too little clinician involvement in AI decisions.
Nearly half of organizations trying to scale AI find it slower and harder than expected, Bain found.
What the clinical research shows
A recent study in the New England Journal of Medicine tested 24 AI systems against 200 de-identified outpatient and emergency department notes. Twenty were general-purpose large language models from seven developers. Four were specialized medical systems.
Priced at 2026 Medicare rates, the general-purpose models' plans cost an average of $314 more per visit than the physician's documented plan. The specialized systems recommended care costing 2.7 to 3.5 times the real-world plan.
Testing and referrals drove most of the gap. In 66.5% of the 200 visits, the physician ordered no diagnostic workup. Every AI system suggested testing in at least some of those cases.
The researchers did not call that waste. Nearly a fifth of the AI plans aligned with clinical guidelines, often catching workups the physician had skipped. "Excess recommendation is not synonymous with waste," they wrote.
Richard Carter, chief technology officer at Proximie, called the finding narrow. The models saw only a written clinical presentation, he noted. They could not question the patient or use the real-time observations a treating physician has. "The finding may be right, but it's narrow, and the lesson is bigger than cost," Carter said.
Patients and the denial fight
Americans are warming to AI in medicine, with conditions. A Gallup survey of more than 5,000 adults, conducted Aug. 12 to 24 with a margin of error of 1.7 points, found 56% of those who saw a provider in the past year are comfortable with AI making diagnosis and treatment decisions if the provider can override it. Only 9% are comfortable with AI deciding without provider review.
A separate survey of more than 1,000 primary care physicians found 36% thought patients would accept override-able AI decisions. Patients said 56%.
Patients are also using AI against insurers. A survey of more than 4,500 working-age adults with private coverage found 21% reported that they or a family member had a coverage denial in the past year. Only about half appealed, according to the Commonwealth Fund researchers behind it. About 43% of those who faced a denial said it led to medical debt, and half of those bills were $1,000 or more.
Tom Whitehead, who has Crohn's disease, said his insurer denied the infusion drug Remicade after he switched plans in 2025 and put him on a biosimilar that worsened his symptoms. He used an AI tool called Claimable to draft an appeal citing federal statutes. He says the drug was approved five days later. That is one patient's account.
What employers face
Relation Insurance advises employers to build AI-driven coding intensity into renewal planning and to press carriers for claims-level visibility. The PwC 9% projection is a forecast from health plans, which have their own interest in explaining premium increases.
It remains unclear whether the roughly $3.7 billion going into healthcare AI produces savings that outweigh the billing and testing increases plans and researchers are flagging. For now, only four in ten health system administrators say they can measure it.
Sources used for this briefing
6 links from 6 independent newsrooms (syndicated copies of the same report count once)
These are the sources it draws on, linked so you can verify.