Last week, the New York Times reported on a new Blue Cross Blue Shield Association analysis that puts a number on something everyone in healthcare finance already feels. Hospitals submitted claims for tens of thousands of patients that described their illnesses as more complex in 2024 and 2025 than in 2023, even though the association found no evidence those patients were treated differently. The association attributes the shift to AI tools that help hospitals capture more patient information and submit larger claims. By its count, more intense care added $942 million in costs for Blue plans over the two-year period. Providers have a response of their own: insurers are using AI to reject their patients' claims.

Both sides have a point. That's the problem.

Health systems are deploying increasingly capable AI to document care and get paid for it. Insurers are deploying their own AI to scrutinize claims and hold back payment. Each side is getting better at what it already does, and patients, employers and taxpayers pay for both.

The industry is building a more sophisticated version of the same adversarial system: more software, more automation and more administrative expense aimed at the same healthcare dollar. For health systems running on thin margins, the arms race carries its own cost. Every new payer algorithm means more denials to work, more appeals to file and more people needed just to collect what was earned.

Eric Larsen has offered a useful way to think about this. He describes the first wave of healthcare AI as inflationary. It has been layered on top of existing people and existing workflows, and it has mostly reinforced the economics of fee-for-service rather than changing them.

We believe the next wave will be different.

It has to be deflationary.

From value transfer to value creation

Health systems will always manage both revenue and cost, and getting paid accurately for the care they deliver matters.

The distinction that matters is between moving money around the healthcare system and fundamentally changing its economics.

The first wave of AI has been very good at helping organizations capture a greater share of the healthcare dollar.

The next wave needs to make that dollar go further.

There's a practical reason to start there. A health system has limited control over what payers will pay. It has far more control over what it spends to deliver care: what it pays for supplies, services and labor, whether vendors honor their contracts, and how much variation shows up across sites and physicians. That is where AI can create value without taking it from someone else.

For a health system, that means using AI to answer much bigger questions. What does it actually cost us to deliver this procedure? What are we really earning after every adjustment and contractual term? Where is margin leaking through price, utilization or variation? Which vendor commitments aren't being met? Which service lines are becoming structurally unsustainable? Where are large teams doing work that software can now do continuously?

And most important: what should we do about it?

Answering those questions gives health systems operating leverage they have never had. And every dollar saved this way is a dollar no one else has to absorb.

The deflationary financial operating system

That's what we're building at Midstream.

Health systems generate enormous amounts of financial, clinical, operational and contractual data. Connecting those datasets has historically required large teams of analysts, consultants and operators. Even then, the answer often arrives months after the chance to act has passed.

AI changes that. It can read cost and revenue together, continuously. It can spot value leakage as it happens, explain why something changed, recommend an action, help carry it out and confirm whether the financial result actually landed.

That last step matters most. Healthcare already has plenty of dashboards pointing to where a problem might be, and it doesn't need AI generating more work for people to review. The job is to remove work, improve decisions and produce financial outcomes a CFO can verify.

Every AI vendor in healthcare should have to answer one question: did total cost go down? A deflationary tool should be held to a deflationary standard.

The next chapter

The first chapter of healthcare AI helped organizations compete harder within the economics of the existing system. The next chapter will change those economics.

Neither side will win the arms race the Times described. Each new algorithm invites a better counter-algorithm, and the cost of the fight keeps climbing. The way out is to make care itself cost less to deliver.

That's the deflationary era of healthcare AI, and it's starting now.