Good morning, readers —

Happy HLTH week to those who celebrate (recap coming next week)! And man, what a week for AI...

Jensen Huang is out at Nvidia after 33 years, and now he's reportedly mounting a comeback. An increasingly Sam Altman-esque situation, though whether he'll actually return remains to be seen.

Meanwhile, healthcare AI had a massive week of its own. Let's get into the top stories:

  • The CMS roadmap to clinical superintelligence

  • Oura's $625M bet on the future of healthcare

  • Google's two big bets on healthcare AI

  • Plus: 61 updates across tools, partnerships, research, policy, and deals, along with AI jobs and more.

Read time: 7 minutes

Our Picks ✨

Highlights if you’ve only got 2 minutes…

1/

The CMS roadmap to clinical superintelligence

CMS released a new roadmap for “Clinical Super Intelligence,” laying out a surprisingly detailed vision for how AI could be regulated, reimbursed, and scaled across healthcare. A few things stand out:

  • AI should be deflationary >> CMS wants clinical AI to lower total cost of care. Software can scale at low marginal cost, move care upstream, catch problems earlier, and reduce unnecessary visits, tests, and hospitalizations.

  • “Better than today” is the bar >> AI should be judged against the healthcare system patients actually have today, including delayed access, fragmented records, uneven quality, and preventable errors. More autonomy would be earned through evidence, staged deployment, and continuous monitoring.

  • FDA regulation could become competency-based >> the roadmap points toward evaluating whether AI is competent for defined clinical tasks, similar to physician credentialing. Post-market monitoring and predetermined change plans could let models continue improving after deployment.

  • Lower-risk AI could reach patients faster >> CMS wants accelerated pathways for wellness and clinical decision support tools, potentially including direct-access diagnostics and treatments for bounded, lower-risk conditions. State AI sandboxes could also provide room to experiment as oversight and liability frameworks evolve.

  • CMS wants to pay for AI that lowers costs >> perhaps the biggest signal. The agency does not want to reward tools that simply generate more visits, clicks, or billing. It wants payment tied to outcomes and demonstrated value.

  • The reimbursement roadmap is getting specific >> CMS points to chronic disease management through ACCESS, preventive screening, AI-enabled accountable primary care, total-cost-of-care models, Medicare and Medicaid reforms, Medicare Advantage, Marketplace products, and state pilots. It even suggests portions of the Medicare Annual Wellness Visit could eventually be delivered through clinical AI.

  • The data rails are already moving >> CMS says its Health Tech Ecosystem has grown from 60+ companies to more than 800, while digital patient access to health data increased from roughly 5% to 60% in one year.

  • Patients remain in control >> the roadmap emphasizes consent, transparency, control of health data, and preserving patient autonomy as AI becomes more capable.

CMS is also clear about what this does not mean. It is not mandating AI adoption, creating a new regulator, or immediately authorizing autonomous care everywhere. The idea is to start with lower-risk use cases, prove they work, and expand autonomy from there.

The economic framing is the biggest signal. CMS is treating AI as a way to expand healthcare supply while lowering the marginal cost of care, then building payment models around those savings.

For health AI builders, this is about as close to a roadmap as you can ask for: improve outcomes, lower costs, earn trust, and get better with real-world evidence. Go forth and build. Accelerate Life. (link)(tweet)

2/

Oura's $625M bet on the future of healthcare

Oura announced plans to acquire Counsel Health for $625M in cash, one of the more interesting frontier health AI acquisitions we've seen. The companies first partnered in May and expanded their relationship in September around CMS's ACCESS model. Oura must have liked what it saw enough to bring the whole thing in-house.

Counsel combines AI-native primary care with board-certified physicians. Patients can start with AI that reviews their history, records, labs, and biometric data, then bring in a physician for diagnoses, prescriptions, referrals, or more complex clinical judgment.

That plugs directly into Oura's 6M+ member base. Oura also disclosed roughly $1.75B in FY2026 revenue, up 92% year over year. Counsel gives it a compelling growth story beyond hardware and subscriptions.

The strategy is interesting because Oura is approaching AI healthcare from the opposite direction of many AI-native care companies. Most start with care or intelligence, then try to pull in medical records, labs, and wearable data. Oura already has years of continuous sleep, activity, recovery, heart rate, and other biometric data.

Counsel lets it build the care layer on top. This has been a missing piece in consumer health: you get insights into your biomarkers or health risks, but still have to find a doctor to interpret them and figure out what to do next. Oura now has a path to closing that loop.

There are several ways this could pay off. Oura and Counsel are participating in CMS's ACCESS model, creating a potential reimbursement pathway for outcomes-based chronic care. Counsel also works with LillyDirect around GLP-1 care, while Eli Lilly invested in Oura earlier this year. (GLP-1s still out-earn OpenAI and Anthropic combined...)

Put it together and Oura could identify a change in your health, understand it using years of data, connect you to care, and measure what happens afterward. That's a much bigger product than a smart ring.

It looks like a strong outcome for both sides. Counsel has agreed to a $625M exit after raising ~$36M, while Oura adds AI, clinical care, reimbursement, and a deeper relationship with its users.

It may also be a preview of where consumer health is headed: start with the data, then build the care system around it.

P.S. We profiled Counsel back in June 2025. Keep an eye on our deep dives. Truffle pig instincts... (link)(linkedin)

3/

Google's two big bets on healthcare AI

Google had a big week in health, launching Gemini for Healthcare for providers and payers, followed by AlphaGenome 2 from DeepMind.

  • Gemini for Healthcare: Google is firmly in the healthcare platform race alongside OpenAI and Anthropic. Its HIPAA-ready agent layer is built for providers and payers, with healthcare-specific skills for chart prep, prior auth, coverage review, appeals, and other workflows.

    The connector list is already deep, spanning Epic, Oracle Health, Optum, TriZetto, CMS data, BMJ, GoodRx, ServiceNow, Microsoft 365, and Google Workspace. Launch partners include HCA, Highmark, Northwell, Optum, Blue Cross Blue Shield of Massachusetts, and ASCO.

    For providers, Gemini can synthesize patient histories, prep charts, and assemble authorization packets. For payers, it can review requests against policy and explain benefits and expected costs to members.

    Epic and Oracle integrations are still read-only today, but the direction is clear. Competition is moving beyond model quality toward connectors, workflows, permissions, distribution, and who can actually get useful work done inside healthcare.

    Google has a compelling hand given its cloud footprint, enterprise distribution, consumer products, Gemini models, healthcare-specific models like MedGemma, and DeepMind research stack. (link)(LinkedIn)

  • AlphaGenome 2: DeepMind also launched AlphaGenome 2, trained on 107M+ individual cell profiles across 68 tissues to predict how genetic variants affect specific cell types.

    On a benchmark involving 28 immune cell types, it nearly doubled the original model's causal-variant identification performance from 16% to 30%, while improving prediction of gene-effect direction from 23% to 43%.

    The bigger promise is connecting genetic variation to actual cell biology. AlphaGenome 2 remains a research tool, but pair models like this with genomic data and eventually clinical decision support, and scalable variant interpretation at the point of care starts to look more plausible.

    You can see the longer-term path: sequence your genome → AI maps disease risks → personalized prevention → identify the biology when disease develops → design a treatment around it → manufacture it in an automated lab.

    We're early, but more of those pieces are starting to come together. (link) (tweet)

Google is building at both ends of healthcare AI: agents that help run healthcare today and models that could change how we understand disease tomorrow. Onwards.

Healthcare AI Rundown 🔍

Latest on tools, partnerships, research and policy…

Tools & Product Updates 🔧

  • Apple and OpenEvidence launch consumer health AI study: Up to 200,000 eligible U.S. participants will test a chatbot, examining how personal health data affects trust, understanding, and usefulness. (link)

  • Abridge adds local care pathways to its Assistant: Responses incorporate health systems' clinical guidance alongside medical literature and patient context, bringing institution-specific recommendations into workflows. (link)

  • Candid launches AgentKits for existing billing systems: Its Denials AgentKit combines rules, AI, integrations, and audit trails. An Eligibility AgentKit is planned for early 2027. (link)

  • R1 expands Phare OS with 6 AI modules: Authorization, audit, utilization management, and denials tools are available, with documentation and recovery capabilities rolling out through 2027. (link)

  • Luminai gives healthcare agents institutional context: Topology organizes operational knowledge for agents. Cleveland Clinic reportedly reduced referral processing from 3+ days to the same day. (link)

  • Lumeris launches Tom Health to extend primary care: Its AI team member supports outreach, coordination, and chronic care between visits, with partners across 14 states, including Ochsner. (link)

  • August AI adds booking and medical bill support: Its free beta agent calls providers, books appointments, compares prescription costs, audits bills, and prepares appeals for physician review. (link)

  • MDCalc launches CrossCover for clinical AI safety: The platform adds privacy safeguards, calculators, and beta evidence checks, launching with AMBOSS and Vera Health. (link)

  • OpenEvidence adds affordability tools at the point of care: Free Access Intelligence tools cover 2,500+ patient support programs, savings offers, enrollment forms, and prior authorization materials. (link)

  • Kin connects visit summaries to personal AI assistants: Patients can authorize read-only access to summaries, notes, and next steps from Claude, ChatGPT, Grok, and compatible agents. (link)

  • Luma launches agents for procedure preparation and follow-up: Agent Launchpad escalates clinical questions to nurses, with company-reported results of 95% patient response and 20+ nursing minutes saved per patient. (link)

  • Ertas releases 2 open-weight medical models: MedErtas supports clinical questions, calculations, and record workflows, with company-reported performance exceeding MedGemma 27B across 11 benchmarks. (link)

  • Concentrate AI launches a gateway for HIPAA-covered inference: Developers can access leading models through one API and BAA, consolidating provider agreements, inference, and billing. (link)

  • Vumedi expands AI search across clinical videos: Physicians can ask questions and jump to relevant timestamps across 100,000+ expert videos and nearly 5,000 podcast episodes. (link)

  • Forus automates prior authorization routing: Germinator identifies coverage, drug-benefit managers, and required forms. In one company-reported case involving 9 plans, submission took 1.3 hours. (link)

  • Blank Bio introduces Ladon for treatment-response prediction: The RNA foundation model analyzes gene expression, with company-reported validation across 140+ oncology tasks. (link)

  • Gather Bio emerges with AI-assisted lupus trial matching: Its platform combines records, patient-reported outcomes, and biomarkers to identify potential participants and deliver consented referrals. (link)

  • Cedars-Sinai releases 3 free research AI tools: ASAREE, Memstara, and GENASYS help researchers build, modify, and monitor agentic systems, aiming to make scientific outputs more transparent. (link)

  • Kaiser physician group launches an AI startup engine: Permanentech will develop clinician-built companies and partnerships. Its first portfolio company uses AI to predict missed cancer infusion appointments. (link)

Partnerships & Deployments 🤝

  • Layer Health + Fairview: Fairview will automate registry abstraction across 10 hospitals and 60+ clinics, starting with stroke, with oncology and cardiology support planned. (link)

  • Ohio State + Microsoft: A 5-year partnership will develop a smart hospital model and apply AI and cloud technology across care, research, education, and operations. (link)

  • HealthEx + Oscar Insurance: Members can connect records through Oscar’s app, giving its Oswell AI agent clinical context for personalized answers, coverage guidance, and care navigation. (link)

  • Glass Health + epocrates: Glass will power broader evidence-based clinical answers within epocrates AI Assist, expanding its conversational capabilities beyond established drug-information resources. (link)

  • Dyania Health + NVIDIA: Synapsis will run inside hospital networks. Cleveland Clinic used the platform to analyze longitudinal records from 7,812 aortic valve replacement patients. (link)

  • Commure + Ontada: Commure is bringing pre-visit summaries, ambient notes, chart search, dictation, and coding suggestions into the iKnowMed oncology EHR, with clinician review retained. (link)

  • HealthEx + Sierra: HealthEx joined Sierra’s Personal Agent Protocol as a design partner, contributing healthcare identity, privacy, and consent requirements to the open standard. (link)

  • Heidi + Samsung Medical Center: The partners will test AI documentation in South Korea, including hospital-record integration and adaptation to Korean clinical standards and workflows. (link)

Research & Evidence 🔬

  • Cognita reports 40% enhanced detection in radiology research: Company-led studies spanning 237,905 exams and 50 conditions found additional expert-confirmed findings. The models remain investigational in the U.S. (link)

  • Physician personas shift AI decisions despite identical facts: Across 20 models and 5,000 cases, persona instructions shifted affirmative responses from 36.9% to 46.4%, without consistent protection from medical fine-tuning. (link)

  • Google evaluates a multi-agent personal health assistant: Researchers used wearable and clinical data from 1,165 participants to assess agents for analysis, health reasoning, and coaching, evaluating system performance rather than clinical outcomes. (link)

  • 85% of surveyed physicians use AI weekly: Doximity’s survey of 3,379 physicians found weekly use increased from 51% last year across clinical and administrative work. (link)

  • Clinician AI use outpaces formal training: Gallup found 89% of surveyed primary care physicians used AI for work, but 20% received training; 56% of patients accepted AI decisions with provider override. (link)

  • Health system leaders report average AI returns of 3.2x: Vizient surveyed 416 leaders across 385 systems, finding a 2.5x median return. Results were self-reported, not independently verified. (link)

  • Only 27% strongly trust automated drug checks without clinicians: Wolters Kluwer found limited consumer confidence, while 99% of surveyed healthcare professionals reported organizational movement toward AI medication workflows. (link)

Policy & Regulation 🏛️

  • AMA opposes licensing AI to practice medicine: Its new policy supports AI as a physician tool while rejecting AI licensure and professional titles, alongside calls for oversight and liability protections. (link)

  • DiMe offers a framework for AI care navigation: Journey maps, an assessment checklist, and opportunity cards help health systems identify care gaps and evaluate appropriate AI uses. (link)

  • 34% of studied AI devices changed after clearance: Researchers tracked 163 FDA 510(k)-cleared devices, finding a median 1.8 years to first update and inconsistent performance-change reporting. (link)

  • Medicaid pressures sharpen scrutiny of AI spending: Reporting on HR 1 finds health systems prioritizing demonstrable returns, including enrollment and administrative tools, while safety-net providers face constrained investment capacity. (link)

  • DOJ alleges AI impersonation cost hospitals nearly $7.8M: Prosecutors say defendants mimicked executives and attorneys to induce fraudulent transfers from 2 hospital systems in Missouri and California. (link)

Deal Desk 💰

Spotlight on latest capital raises, M&A, and investments…

FUNDING

  • Vumedi, an Oakland, Calif.-based peer-to-peer learning community for physicians, raised $104M in Series B funding from Oak HC/FT. (link)

  • Charta Health, a San Francisco-based AI medical chart review platform for billing accuracy, raised $55M in Series B funding. Insight Partners led, joined by Madrona and Bain Capital Ventures. (link)

  • Mosaic Intelligence Labs, a San Francisco-based surgical intelligence and robotics company, emerged from stealth with $47M in seed funding backed by 8VC, Fundomo, Lingotto Innovation, and Obvious Ventures. (link)

  • Knit Health, a company building a healthcare world model, raised $35M in Series A funding led by NEA. (link)

  • Orakl Oncology, a Paris-based cancer outcome prediction platform, raised $34M in Series A funding co-led by Norrsken VC and Singular. (link)

  • Zus Health, a Boston-based connected healthcare data platform, raised $33M in Series B funding. Drive Capital led, joined by Sanofi Ventures and HealthX Ventures. (link)

  • Duet, a network of nurse practitioner-owned primary care practices, raised $15M in Series A funding led by Kairos, bringing total funding to $22.5M. (link) 

  • XpertDox, an AI-first healthcare company, today announced a $20M Series A investment round led by Savant Growth. (link)

  • Optain, an AI-powered retinal imaging and screening platform, raised an additional $13M, expanding its Series A to $39M to scale retinal screening across health systems. (link)

  • Alterra Labs, a New York-based maker of simulated medical environments to train and test AI models, raised $12.5M in seed funding. Oak HC/FT led, joined by Flare Capital Partners and Company Ventures. (link)

  • Sleep.ai, a Carlsbad-based sleep intelligence platform that turns phone and wearable data into personalized guidance, raised $5.3M in new funding. (link)

  • Avanlee Care, a Montana-founded family caregiving technology company, raised $3.5M, bringing total funding to $6M. BEA Venture Fund and Fairbridge VC participated. (link)

M&A

  • Oura + Counsel Health: Oura plans to acquire the AI virtual care company for $625M to bring records, informed medical AI, and physician care into its platform. (link)

  • Elligint Health + Laguna Health: Elligint acquired the conversational AI company to add AI for health plan care management, and Laguna will keep operating under its own name as a subsidiary. (link)

  • Commure + Elaborate: Commure acquired Elaborate, an AI-powered clinical intelligence platform that helps reduce inbox burden by interpreting results and generating patient-friendly follow-up messages inside the EHR. (link)

  • Wolters Kluwer + AgileMD: Wolters Kluwer agreed to acquire AgileMD for $40M upfront and up to $60M in earnouts, expanding UpToDate Expert AI further into clinical decision support workflows. (link)

  • Inovalon + DrFirst: Inovalon agreed to acquire DrFirst, adding its e-prescribing and medication management capabilities to Inovalon’s healthcare data and analytics platform. (link)

GRANTS

  • AWS, Amazon’s cloud computing arm, committed $100M of AI and cloud technology through its Health Impact Initiative. The program helps health organizations build solutions for underserved communities. (link)

  • CareSource, a nonprofit health insurer, launched a $1M, two-year grant program supporting AI solutions for Medicaid care coordination, member engagement, behavioral health, and access to care. (link)

as of 11/16/26

Other Relevant News 🔍

News, podcasts, blogs, tweets, resources, etc…

  • AI could shift more specialist care to generalists (link) 

  • AI diagnoses a rare condition doctors missed for decades (link)

  • McKinsey expects AI to shift healthcare growth toward software and data (link)

  • Big Pharma can rival big software (link)

AI Job Opportunities 💼

Explore our AI Job Board or contact us to feature roles in our newsletter…

  • Clinical Project Lead (MD) at Lumos, health AI evaluation company
    $180K - $200K | SF (link)

  • Member of Clinical Staff at Kinetic Systems, healthcare AI research lab

    $150K - $250K | SF (link)

  • ML Systems Engineer at Periodic Labs, AI and physical sciences company
    $250K - $350K | Menlo Park, CA (link)

Visuals of the Week 📸

Funny memes, cool pics, and interesting data from around the web…

That’s it for this week friends! Back to reading — I’ll see you next week.

Stay classy,

— Healthcare AI Guy (X/Twitter | LinkedIn)

PS. I write this newsletter for you. So if you have any suggestions or questions, feel free to reply to this email and let me know

How was this week's newsletter? Tap your choice below👇

Login or Subscribe to participate


You may also like these

Read all
arrow-right