Value Prompt · by Nico Neugebauer
A weekly email on AI in healthcare
Every Tuesday morning I send you the 3 or 4 AI and healthcare stories I'd want you to know about, what each one changes for your work and the tools worth keeping. I find them with a scout I built, which reads 107 sources, most of them every night.
It's free, it comes every Tuesday morning and you can unsubscribe with one click.
The first issue arrives Tuesday, October 20.
Value Prompt
Sample issue · week of October 5
01The news that mattered
Utah will let AI write prescriptions, one phase at a time.
The so what
If you want AI that acts on its own in healthcare, the way in now is a state sandbox, one phase at a time, and each phase needs evidence before it opens.
02The Toolbox
Heard: the AI says back what it heard before it starts the work
03Only for subscribers
The 107 sources my scout reads
If you build with AI and don't write code, the Toolbox is for you too.
3 or 4 stories, each with what it changes for the people who run care, pay for it or build for it.
One tool a week for people who don't write code: what I built or tested, whether it held up, and the link so you can use it yourself.
A tool, the numbers behind a post or a list like the 107 sources my scout reads.
My scout reads medical journals, regulators, the healthcare trade press, the AI labs and the people building with AI, most of them every night.
It groups what it finds into stories, leaves out paid promotion and news with little in it, and scores each story on whether it changes what someone working in healthcare should do, believe or try.
Every story links to its source and gets checked against it, I flag any number that comes from a company grading itself, and every tool is one I built or tested.
You get the 3 or 4 that matter and what each one changes for your work.
My scout finds the stories, and I choose which ones make it, edit every issue and put my name on it.
From a sample issue, the week of October 5
Utah's Office of Artificial Intelligence Policy approved 3 new pilots in its AI sandbox, including Nolla Health, whose app assesses mild-to-moderate acne from photos and can prescribe from a short list of topical drugs. At first, 2 physicians review every prescription before it goes out. In later phases they review after the fact, and each company needs the state's written approval to get there (Healthcare Dive). None of the pilots has launched yet, and Sergei Polevikov at AI Health Uncut points out that Nolla's 96% accuracy claim comes with no published study (AI Health Uncut).
The so what
If you want AI that acts on its own in healthcare, the way in now is a state sandbox, one phase at a time, and each phase needs evidence before it opens. So if you're building anything clinical, the study and the record of who signed off are part of the product from day one.
Nico Neugebauer
AI product & growth leader
I'm Nico Neugebauer, Director of AI Strategy & Transformation, and I shape the AI strategy for a value-based care provider enablement portfolio. Before that I advised health systems on strategy and finance at Kaufman Hall (now part of Vizient), then sold multi-year value-based care partnerships at Lumeris, including ACO REACH and Medicare Advantage risk arrangements. I co-founded Wallace Stories, where 2 people and an AI pipeline I built with Claude Code turn a family's memories into a printed children's book, and building it taught me where AI holds up in production and where it quietly breaks. I don't write code myself (Claude Code does), so you can set up every tool I hand you without an engineer.
When a number comes from a company grading itself, or a claim has no published study behind it, I say so and tell you which number I'd keep, if there is one. No company can pay to change what I tell you.