Time, money and mojo for people who take call and want to understand now and what comes next.
New here? Every Saturday a bunch of AI takes the week’s best idea in each of three buckets and cut it down to something you can read between patients.
AI reads the week so you don’t have to. I add / subtract / comment on a few things.
Obviously, these are not the views of any big org and are mainly AI generated, so take with a grain of salt.
But it is what super intelligence is thinking about our future.
Let’s go.
— Kevin
1. TIME
The AI didn’t take the abstractor’s hours. It moved them.
On Wednesday, Layer Health announced an enterprise deal with Fairview Health Services to automate clinical registry abstraction across its Minnesota hospitals and clinics.
Stroke goes first, through the American Heart Association’s Get With The Guidelines–Stroke program. Then, once the tool is checked against Fairview’s own historical charts, the cancer and cardiology registries.
If you’ve never sat with an abstractor: they read hundreds of pages across many encounters to fill in one registry form. Was the last-known-well time documented? Was the statin ordered at discharge? Slow, careful, invisible work.
The tool pre-fills each field and highlights the sentence in the chart it came from. The abstractor checks it, fixes it, signs it. HIT Consultant’s write-up calls this “keeping final clinical discretion in human hands.”
David Sontag, Layer’s CEO and co-founder, put the goal as teams that “spend less time searching for answers and more time moving quality improvement initiatives forward.”
Here’s the turn. We talk about AI saving time as if the minutes vanish. They don’t. They move. Out of searching, into deciding.
The vendor claims more than 70% time savings per chart at peer health systems. That’s their number, not Fairview’s yet. But even if it holds, the abstractor isn’t doing less of the job. She’s doing more of the part only she can do.
So look at your own week. Which hours are searching, and which are deciding?
The first kind is what a tool should take. The second kind is the job. Guard it.
A win.
(Words: 255. A minute.)
“Our platform brings clinically intelligent machine learning to a workflow that mirrors how abstractors already work.” — David Sontag, Layer Health
Dig deeper: the Fairview announcement. Note the sequence: validate on your own old charts first, then expand.
2. MONEY
Medicare finally gave software its own line. Then it priced the service, not the software.
On Wednesday, CMS released its 2027 hospital outpatient final rule. Inside is a new payment class: Software as a Medical Service, or SaMS. It gets its own status indicator, O1, which means it’s paid separately instead of disappearing into the visit. It starts January 1.
The rule text admits the problem. Medicare’s payment systems were built to pay for “material resources,” not for “proprietary algorithms and scalable, non-material costs.”
Now look at one code. Vincent Capone‘s company, Spectral AI, got FDA De Novo authorization in May for DeepView, which uses imaging and AI to predict whether a burn will heal. Its code, CPT 0972T, was status E1: not covered at all.
A commenter asked CMS to make it a SaMS and pay $2,750.50, “stating that this would cover the cost of the software.” CMS agreed it’s a SaMS. Then it crosswalked to a cardiac-risk software code and set the rate at $250.50.
Here’s the turn. The win is real, but it isn’t the rate. CMS priced the clinical service the software supports. Your license isn’t on the schedule. The hospital’s margin is.
And CMS declined to include non-device CDS. No FDA authorization, no SaMS payment.
📜 History rhymes: When Medicare moved hospitals to fixed DRG payments in 1983, hospitals stopped asking what care cost and started asking what the category paid. These interim bands are categories.
😤 “$250 is a rounding error.” It’s also the first row with software’s name on it.
What does your code’s band pay, minus what each use costs the hospital?
A win.
(Words: 273. A minute.)
“…primarily designed to pay for services that rely on material resources, rather than technologies whose value is driven by proprietary algorithms and scalable, non-material costs.” — CMS, CY 2027 OPPS final rule (CMS-1850-F)
Dig deeper: Reed Smith’s summary by Nancy Bonifant Halstead and Allie Klimkiewicz. It spells out the FDA requirement and why CMS kept rates close to 2026.
3. MOJO
Nobody asked a trauma surgeon to read Congress. He did it anyway.
Christian Péan, MD is an orthopaedic trauma surgeon at Duke. He’s also an associate residency program director, core faculty at the Duke-Margolis Institute for Health Policy, and a co-founder of RevelAi Health.
That’s a full plate.
This week, right after the midterms, he published Where Congress Stands on Health AI. It’s a record of 66 members of Congress and senators-elect, built from 179 bills, letters, votes and statements since 2023. Each one is dated and linked to its source.
A few things it shows. 41 members have a health AI record. 25 lean toward guardrails, 13 toward adoption and payment. 33 members have gone on the record about WISeR, Medicare’s AI prior-authorization model.
His summary is the line I keep thinking about: health AI splits Congress by purpose more than by party.
Here’s the turn. The obvious read is “a surgeon built a policy tracker.” The real one: the people with the most at stake in health AI policy are the ones with the least time to follow it. So one of them did the reading for everyone else.
The animated timeline is free. The deeper explorer is paid. The work is the same either way: somebody noticed a gap clinicians keep tripping over and filled it on his own time.
That’s mojo. No grant. No mandate. Just the unasked-for job.
You say you want a say in what comes next. Do you know where your own representative stands on WISeR?
Ten minutes on his timeline would tell you. Who else needs the link by Monday?
A win.
(Words: 268. A minute.)
“Health AI divides Congress by purpose more than by party.” — Christian Péan, MD
Dig deeper: the record itself. Start with the WISeR records. That’s the policy that touches your patients first. [His tracker is beautiful BTW]
🔭 BONUS: THE ADJACENT POSSIBLE
One thing from outside health that’s about to be reimagined inside it.
On Allstate’s third-quarter earnings call this week, CEO Tom Wilson said something you rarely hear from an incumbent: “We will be available to all personal agents.”
He means the AI assistants consumers are starting to carry. Instead of fighting them, Allstate plans to be the company those agents can reach, quote and buy from. Its customer-service assistant, launched in July, already handles 15% of chat volume. Next year it plans its own protection-focused personal agent.
Outside health, that’s a sales channel.
Inside health, it’s a preview. Your patients’ agents will soon try to book the visit, refill the prescription, check the price, ask about the result. Every health system will face Allstate’s choice: be reachable on its own terms, with a clear record of who authorized what, or be the portal the agent gives up on.
One more line worth stealing. Wilson said Allstate has 25% fewer people in technology than four years ago, “and that’s not just using AI.” A CEO not crediting AI with everything. Rare.
The caveat is big: an agent acting for a patient raises consent, identity and liability questions insurance never had to answer.
Dig deeper: the CIO Dive piece. Read it for what “available” requires: an interface, terms, and a way to complete the transaction without a human.
Tell me your thoughts: kevin@clinicians.build.
— AI & Kevin
(please verify content for yourself, partially AI generated and may contain errors)





