Sponsored by: Trustible
What do 136 AI incidents reveal about AI risk in health systems?
Trustible’s new report sorts publicly reported AI incidents impacting health systems from the AI Incident Database into 20 categories, each with suggested mitigations. It spans clinical care, patients, operations, and the back office. Use it to support safer AI for your community, train your teams, and stay informed. Get the free report here!
Welcome back, readers!
Hope everyone enjoyed the long weekend last week! Our team was off, but healthcare AI definitely didn't slow down. Here's what we've got for you from the past two weeks:
AI is doing exactly what healthcare pays it to do
Google tests its AI with real patients
The $1.8B push to build a virtual cell
65 tools, partnerships & deals, plus new research, policy updates, AI jobs & link-worthy content
Read time: 7 minutes
Our Picks ✨
Highlights if you’ve only got 2 minutes…
1/
AI is doing exactly what healthcare pays it to do
A new PHTI report makes the case that AI is exposing the incentives already baked into healthcare. Right now, that means a lot of the fastest adoption is happening around documentation, coding, claims, and prior auth, where more activity can translate directly into more revenue or lower administrative cost. A few things stood out:
AI scribes can increase billing intensity. One multihospital system saw Level 5 visits rise 5% and Level 4 visits 7% after deploying a scribe, adding about $1,004 in monthly revenue per provider.
Prior auth is getting faster, not necessarily cheaper. Providers and payers are both automating submissions, denials, and appeals, but PHTI found no evidence yet that AI lowers average cost per claim once the tool itself is included.
The higher-value clinical use cases are still stuck. Hypertension management and mental health are two areas where AI could expand access and deliver more continuous care, but payment, liability, and licensure still get in the way.
PHTI’s proposed fix is mostly about incentives: better real-world evidence, clearer rules for coverage and liability, payment tied more closely to outcomes, and redesigning workflows instead of layering AI on top of broken ones. CMS’s new ACCESS model, with 160+ organizations, is one early example.
AI is a multiplier, and healthcare needs to make sure it is multiplying the right things. If we keep rewarding more coding, more claims, more denials, and more documentation, AI will get extremely good at producing exactly that.
Healthcare AI probably needs its own version of value-based care, where better outcomes, lower costs, and broader access are what get rewarded. Otherwise, we may just end up making an inefficient system much better at being inefficient. (link)

2/
Google tests its AI with real patients
Google just published a new clinical study of AMIE, its conversational medical AI system, showing what this kind of tool could look like inside an actual patient visit.
AMIE can take a patient history, ask follow-up questions, and prepare a clinical summary, differential diagnosis, and management ideas for a doctor to review before the appointment. In this study, 114 urgent-care patients at Beth Israel Deaconess used AMIE, with 98 completing both the AI interaction and their PCP visit. The peer-reviewed results were published in The Lancet. Some highlights:
Zero conversations were stopped for safety reasons. Supervisors noted one hallucination across the study.
AMIE included the eventual diagnosis in its top 7 possibilities 90% of the time, and in its top 3 in 75%.
Of the PCPs who reviewed AMIE before the visit, 33 of 44 said it helped them prepare, while 25 said it may have changed their approach.
Clinical evaluators rated the conversations favorably across 87% to 100% of assessed criteria.
Having AI take a patient’s history before the visit so doctors can spend less time gathering information and more time actually caring for patients is just a good idea. More of this please. (link)(linkedin)

3/
The $1.8B push to build a virtual cell
A pretty ambitious AI biology project is taking shape. Biohub, the US government, Google DeepMind, Isomorphic Labs and Meta are teaming up on a $1.8B effort to build the data layer for a “universal virtual cell.”
The bet is that with enough real biological data, AI can start predicting how a cell will respond to a drug, genetic change, or other intervention before scientists spend the time and money testing it in the lab. Here’s who’s committing what:
Biohub: $500M toward imaging, measurement technology, data generation and external research.
Department of Energy: $500M+ over five years for biological measurement, modeling, compute and national lab infrastructure.
NIH: $500M+ in federally funded datasets, repositories and other research resources.
Google DeepMind + Isomorphic Labs + Meta: another $300M toward the Virtual Biology Initiative.
The scale of the data problem is wild. Biohub says accurate models may eventually need observations from trillions of cells, compared with hundreds of millions in today’s largest datasets. Commercial partners will get one year of exclusive access to the data they help generate before it becomes public.
This reads like an attempt to build the infrastructure for predictive biology. If it works, researchers could test far more ideas virtually, save expensive wet-lab experiments for the most promising ones, and eventually model disease at the cellular level. Very cool. (link)(linkedin)

Healthcare AI Rundown 🔍
Latest on tools, partnerships, policy and research…
Tools & Product Updates 🔧
Abridge launches an AI assistant inside clinical workflows: Clinicians can ask questions, surface chart context and guidelines, fill calculators, and draft referrals and handoffs for review. (link)
Teladoc adds an AI scribe and camera-based vitals to Solo: SoloScribe is available for ambient documentation, while SoloVitals’ smartphone-based heart and respiratory rate measurements remain in private preview. (link)
JAMA is launching an AI journal: JAMA AI, the network’s 14th journal, will debut in early 2027 under editor Roy Perlis, covering AI and computational technologies across healthcare. (link)
Cellino lets adults bank their immune cells for future use: Reservoir preserves cells from a blood draw for long-term storage and releases them at the participant’s direction. (link)
Happy Health reports FDA clearance for cuff-free blood pressure: The company says Happy Ring enables passive day-and-night measurements without a traditional cuff or external calibration. (link)
Elation connects primary care billing from note to payment: Automated payment posting is in beta, with new-customer access Nov. 1 and complex or disputed claims routed to staff. (link)
Tandem Health launches sovereign AI for European healthcare: Sovereign Health keeps both patient data storage and AI processing on European-owned cloud infrastructure, giving providers control over where processing occurs. (link)
Infinitus launches AI agents for pharma-provider outreach: The agents automate repetitive provider communications for pharmaceutical sales and reimbursement teams, extending the company’s automation tools into pharma engagement workflows. (link)
Gluware automates patching for connected medical devices: Developed with Ohio State Wexner Medical Center, the platform matches vulnerabilities to applicable fixes and routes remediation through hospital approval workflows. (link)
Arcadia launches 3 healthcare platforms with embedded AI: Ascent, CareJourney and LifeScience combine its data infrastructure with AI workflows for clinical performance, market intelligence and real-world evidence. (link)
Microsoft’s Quine ranks pancreatic cancer compounds over a weekend: Built with the Broad Institute, the biology research system generated predictions that subsequent lab experiments supported. (link)
Danaher builds an AI lab for antibody discovery: The facility will combine robotics and AI, targeting up to 8x faster discovery with scaled operations expected in early 2027. (link)
Oura earns medical-device quality certification for its software: ISO 13485 certification covers its quality management system, including Fertile Window software; the certification itself does not constitute FDA product clearance. (link)
Google launches $180 Fitbit Edge with AI health tracking: HealthGuardian surfaces blood pressure and insulin-resistance trends alongside heart rate alerts, with the features intended for general wellness. (link)
HHS awards $1.5M for clinical AI agents and lab interoperability: MedStar will develop patient-controlled AI workflows, while Regenstrief targets gaps in standardized laboratory data exchange. (link)
AHRQ launches a $7.95M healthcare AI challenge: The initiative will fund competitions focused on speeding clinical evidence reviews and finding better ways to use federal healthcare datasets. (link)
UW protein design institute gets 7M hours of AI compute: A Huang Foundation grant will support open-source AI tools for designing proteins and potential new medicines. (link)
Anthropic launches Haiku 5.5 for high-volume AI workloads: The small model targets summarization, classification, browser use and live support, with a focus on speed and lower cost. (link)
Meta launches Muse Wonder with persistent memory: The new model targets tasks across healthcare, law and finance, with memory designed to retain user context over time. (link)
Partnerships & Deployments 🤝
AstraZeneca + insitro: AstraZeneca partnered with insitro on AI-enabled drug discovery for liver disease and cardiomyopathies, with $40M upfront and up to approximately $1.46B in potential milestones. (link)
University of Texas System + Abridge: The University of Texas System expanded Abridge across its enterprise, bringing AI-powered clinical documentation to a health system handling nearly 11.7M outpatient visits annually. (link)
UMass Memorial + Microsoft + Accenture: UMass Memorial partnered with Microsoft and Accenture to unify EHR data on Azure and deploy AI-powered clinical and operational workflows across the health system. (link)
Northwestern Medicine + OpenEvidence: Northwestern Medicine partnered with OpenEvidence to embed cited clinical answers directly into EHR workflows, with single sign-on access for physicians, nurses, and pharmacists. (link)
Johns Hopkins Medicine + OpenEvidence: Johns Hopkins partnered with OpenEvidence to develop clinical AI tools combining patient history with medical evidence and create education for clinicians practicing with AI. (link)
Verily + ARPA-H: ARPA-H selected Verily to build the shared data platform for its CIRCLE program, supporting AI research into immune responses and interventions for critically ill patients. (link)
Tempus + Moderna + Merck: Tempus partnered with Moderna and Merck to provide sample collection and sequencing supporting the potential commercialization of personalized cancer therapy intismeran autogene. (link)
Nextech + Assort Health + EliseAI: Nextech partnered with Assort Health and EliseAI to integrate agentic AI for scheduling, intake, communications, and follow-up into specialty care workflows. (link)
Banner Health + ElevenLabs: Banner Health deployed ElevenLabs voice AI agents to handle patient calls and appointment scheduling, escalating conversations to human teams when needed. (link)
Intermountain Health + Microsoft: Intermountain Health deployed Microsoft’s SmartImpression to help radiologists draft report impressions for review, with roughly 75% adoption as the system transitions to PowerScribe One. (link)
Mayo Clinic + Microsoft: Mayo Clinic and Microsoft worked with nurses to identify and develop five clinical AI agent concepts from more than 450 submissions, spanning paperwork automation and patient guidance. (link)
University Hospitals + ESO: University Hospitals expanded its partnership with ESO to add outcomes feedback and prearrival predictions across an EMS network connecting 21 hospitals and 300+ agencies. (link)
Silna + Forus: Silna partnered with Forus to automate insurance workflows, including benefits verification across more than 1,000 payers. (link)
Ensemble + Penelope Health: Ensemble partnered with Penelope Health to bring real-time payer policies into revenue cycle workflows, identifying coverage and documentation requirements before care begins. (link)
Commure + Compassus: Compassus expanded Commure’s AI referral automation across 79 programs in 19 states, processing 500,000+ annualized referrals and cutting referral decision times by 64%. (link)
Novo Nordisk + Databricks: Novo Nordisk built its FounData platform on Databricks to unify development data and AI workflows, with adoption reaching more than 20% of its development organization. (link)
Verana Health + Bitfount: Verana Health partnered with Bitfount to combine AI analysis of ophthalmic imaging and EHR data to identify potentially eligible patients for clinical trials. (link)
Butterfly Network + Bridge Neurotech: Butterfly Network partnered with Bridge Neurotech to use its Ultrasound-on-Chip technology in developing a fully noninvasive brain-computer interface without an implanted device. (link)
AWS + 11 Pediatric Health Organizations: AWS awarded more than $1.8M to 11 organizations developing cloud and AI projects across rare disease, epilepsy, cancer, and treatment adherence. (link)
Suki + Wolters Kluwer: Suki partnered with Wolters Kluwer to integrate UpToDate Expert AI into its new Ask feature, letting clinicians query medical evidence directly within existing workflows. (link)
Research & Evidence 🔬
OpenEvidence handled 271M queries from 952,000+ U.S. clinicians: A JAMA study spanning 2.5 years found treatment questions outpaced diagnostic questions; OpenEvidence employees coauthored the analysis. (link)
Mental health AI pilot reports lower depression and anxiety symptoms: A single-arm preprint reported improvements among 299 adults over 10 weeks, without a control group. (link)
42% of surveyed patients used AI instead of a doctor: Cost drove the substitution in PayZen’s survey of 1,000 U.S. adults with recent surgery or hospital visits. (link)
Google’s MedGemma medical models land in Nature Medicine: The open models improved chest X-ray classification and medical image question answering compared with their base models, according to the paper. (link)
OncologyBench puts cancer AI agents through 25 multistep tests: Latch Bio’s benchmark evaluates how agents weigh evidence and make decisions across clinical and oncology research scenarios. (link)
DrugTargetWorld tests AI scientists on simulated drug discovery: The preprint uses synthetic biobanks with known causal relationships to evaluate whether agents can identify promising drug targets. (link)
BioinvestGPT says virtual trials predicted 5 of 6 drug outcomes: The company uses AI-generated clinical trial simulations to forecast whether experimental medicines will succeed in human studies. (link)
Human brain organoids help a computer recognize Braille: A decoder classified signals from 3 lab-grown neural clusters with 83% accuracy, compared with 61% from 1. (link)
Policy & Regulation 🏛️
Major healthcare groups join state oversight effort for medical AI: AHA, AMA, Anthropic and Elevance joined an FSMB workgroup developing recommendations for state medical boards overseeing clinical AI. (link)
AMA outlines a governance framework for autonomous clinical AI: Clinical authority should be earned through evidence and maintained through performance, with defined oversight, permitted actions and accountability. (link)
FDA plans new guidance for medical AI in 2027: Its agenda includes evidence recommendations for generative AI mental-health devices and updated software oversight policies; the documents remain planned. (link)
Bipartisan bill targets hacking by autonomous AI agents: The AI Agent Accountability Act would create potential civil and criminal liability when developers or operators deploy agents that illegally access computer systems. (link)
AHA pushes back on BlueCross’s $942M AI coding claim: The hospital group disputes the link between AI billing and higher coding intensity, arguing sicker, more complex patients help explain the trend. (link)
Deal Desk 💰
Spotlight on latest capital raises, M&A, and investments…
FUNDING
General Medicine, an AI-powered healthcare store, raised $120M in Series B funding. Andreessen Horowitz led, joined by Eli Lilly, Mercy Health, and others (link)
Suki, an AI ambient scribe for clinicians, raised $90M in Series E funding. Healthier Capital led. (link)
Abby Care, a San Francisco-based AI platform that pays family caregivers, raised $82M in Series C funding. Transformation Capital led, joined by CVS Health Ventures and others. (link)
Multiply Labs, raised $75M in Series B funding led by Dr. Patrick Soon-Shiong with NantWorks, with AstraZeneca, Lingotto, and Teradyne also participating. (link)
Sabi, a Palo Alto brain-AI interface maker, raised $50M in seed funding. Khosla Ventures led, joined by Accel and others. (link)
Scarlet, a London-based AI medical device certifier, raised $50M in Series B funding. Headline led, joined by Atomico and others. (link)
Healthleap, an AI clinical screening platform, raised $38M in seed and Series A funding from Sequoia Capital, First Round Capital, and Hummingbird Ventures. (link)
Vitalize, an AI-native platform for health system labor and capacity, raised $31M in Series A funding. Oak HC/FT led, joined by Norwest and others. (link)
Allara, an NYC-based women's health startup, raised $30M from Index Ventures, GV, Camino Partners, Foreground, and Family Fund. (link)
WhiteLab Genomics, an AI genomic medicine company, raised $26M in Series B funding led by AVP to advance its AI-designed therapies and expand globally. (link)
Brellium, a New York-based AI clinical compliance platform, raised $23M to give providers audit-ready records. Left Lane Capital and First Round Capital led, joined by Digital Health Venture Partners. (link)
Vocca, an AI voice platform for patient calls, raised $20M in Series A funding. Norrsken VC led, joined by Heal Capital, Speedinvest, and Firstminute Capital. (link)
Iris Labs, a Flagship Pioneering company building personal AI products for psychological wellbeing, launched with an initial $20M commitment from Flagship. (link)
Confido Health, an agentic AI operating system for medical practices, raised $12.5M in Series A extension funding, following its $10M Series A. Z47 led, joined by Justin Saks, Together Fund, and Blume Ventures. (link)
Readily, a San Francisco-based AI compliance platform raised $11M in Series A funding. Define Ventures led the round. (link)
Harbor, a San Francisco-based provider of risk assessment and insurance for AI agents, raised $10M in seed funding from Base10 Partners, Abstract, 8VC, and others. (link)
Otter Health, an AI clinical quality platform, raised $8M in seed funding. Fika Ventures and Evidenced Capital co-led. (link)
Onyx Bio, a Berkeley-based company building a virtual immune system, raised $8M in seed funding. Dimension Capital led, joined by Haystack and Refactor Capital. (link)
Outro, a medication tapering platform, raised $7M in pre-seed and seed funding from Listen Ventures, Cake Ventures, LAUNCH, and others. (link)
Tenix Health, a New York-based AI financial platform for health systems, raised $4M in seed funding. Jeff Horing (Insight Partners) led, joined by Refract Ventures. (link)
Intake AI, a San Francisco-based company building AI agents for clinical trial startup and operations, raised $2.5M in seed funding. (link)
Doctolib, a French healthcare scheduling and AI platform, will receive a new stake from NVIDIA as part of its $2B European AI push. (link)
M&A
Agilent + Proscia: Agilent agreed to acquire the digital pathology software provider for $240M in cash, plus up to $30M in milestones, adding AI-enabled pathology software to its instruments and reagents. (link)
Zocdoc + Sohar Health: Zocdoc acquired the insurance eligibility automation company, adding coverage and benefits verification to its booking platform and partner network. (link)
iRhythm + VitalConnect: iRhythm completed its $287.5M acquisition of the wearable biosensor company, adding VitalConnect's AI-enabled cardiac monitoring algorithms and mobile telemetry to its Zio platform. (link)
Artisight + TMG Global: The deal brings healthcare infrastructure and deployment expertise in-house to help scale smart hospital rollouts across U.S. health systems. (link)
Other Relevant News 🔍
News, podcasts, blogs, tweets, resources, etc…
AI Job Opportunities 💼
Explore our AI Job Board or contact us to feature roles in our newsletter…
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,
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👇



