What you get every issue:
Three curated healthcare AI stories with my take. Companies doing the work.
Skills and techniques worth knowing. And opportunities worth pursuing. Just signal.
Three things worth your time this week:
1. GOVERNANCE JUST GOT A CREDENTIAL
The Joint Commission launched its Responsible Use of AI in Healthcare (RUAIH) certification — the first built specifically for healthcare organizations. A voluntary stamp for orgs that "use AI responsibly."
It certifies the organization, not the product — governance, data
management, bias reduction, monitoring, and training. That's the deployment gap written into a checklist. My take: When the connective tissue becomes a credential, the people who can build it become the hire. This is a tailwind, not a hoop. Read it:
2. A MODEL GOT SWITCHED OFF IN DAYS
A US government export directive suspended all access to two frontier Claude models. Why it matters: If one of them sat inside your clinical product, you just learned what "rented land" actually costs.
Every team treats the model as the durable part and the plumbing as disposable. It's the reverse. Models are the volatile layer.
My take: Single-model dependence is now a documented operational risk, not a slide in my book. Multi-model fallback stopped being optional this week. Full story:
3. THE TIME-SAVINGS NUMBER IS REAL NOW
Philips' Future Health Index 2026 surveyed 2,000+ clinicians across 10 countries. Nearly two-thirds increased AI use; 46% report saving 132+ hours a year. But here's the nuance: That's three working weeks recovered — and it's measured, not anecdotal. The productivity case is settled. The deployment case is not. Why this matters for builders: The question is no longer "does it help?" It's "can you ship it, govern it, and get paid for it?" That's the whole game. Source:
COMPANIES DOING THE WORK
Hetairos — Predicts 102 CNS tumor subtypes from routine slides in 12 minutes.
Medaica — Put an AI-assisted cardiac exam into a home kit; validation is the open question.
Garner Health — $100M Series E for AI-assisted care navigation.
AI TOOLS WORTH KNOWING
Claude for Healthcare — HIPAA-ready connectors (CMS, ICD-10, NPI, PubMed) — turns a frontier model into a prior-auth and reporting agent. (Production)
Medicomp (MCP) — Uses Model Context Protocol to wire LLMs to trusted clinical sources — the integration standard the deployment gap lives in. (Announced)
M1 + AI kit (Medaica) — Home cardiac capture with AI review — shipping ahead of its evidence base. (Production)
WORK & PARTNERSHIP OPPORTUNITIES
AI Governance / RUAIH program leads — Health systems (multi) — New cert, new roles.
Head of Amazon Health Services — Amazon — Leadership reset signals fresh hiring.
RPM / cardiac validation roles — Medaica — Product shipped, evidence still needed.
WHAT I'M WORKING ON:
I'll be sending of the State of Health AI 2026.
I've been working on an internal tool that I'll release to the public:
It's is a curated marketplace for hiring AI agents, humans, or human-and-agent pairs to run a single healthcare or pharma workflow — priced on the outcome, not the hours or the tokens.
If you've built an agent or run a specialist service for healthcare ops, agentslist puts you in front of buyers searching by the exact workflow you solve — and it forces the framing that actually closes healthcare deals: name the outcome, price the outcome. It's curated, so you're not buried in a feed of 500 generic 'AI tools.'
Tell me your workflow and pricing and I'll get you listed.
Writing the chapter on model resilience — and this week handed me the case study. I changed the title from the Deployment Gap to The Last Mile: Turning AI models into medicine people actually use. This is a working kit for getting clinical and pharma AI into production — and keeping it there.
— Victor
Reply to this email. I read every message.
Yuuki Edge • Healthcare AI Intelligence •