Sponsored by: GenHealth.ai
Healthcare organizations burn thousands of staff hours a year on eligibility checks, prior auths, and claims follow-up, spread across phone calls, faxes, the EHR, and payer portals.
GenHealth's AI agents run those processes end to end, executing on over 20 million actions per year, inside the systems groups already use. Piedmont Medical Solutions grew paid-to-date collections 34.2% after switching.
Book a call with us to see what your collections number could look like.
Hey readers —
Hope you had a great Labor Day weekend! We’re back with your weekly roundup of the most important health AI news from the past 7 days:
OpenAI is getting more serious about healthcare
OpenEvidence is going deeper into specialist AI
Oura gives us a consumer health pure play to watch
30 new tools/partnerships, 8 funding updates, new AI jobs & link-worthy content
Read time: 6 minutes
Our Picks ✨
Highlights if you’ve only got 2 minutes…
1/
OpenAI is getting more serious about healthcare
OpenAI had a busy week in health. ChatGPT for Healthcare (their enterprise offering) can now pull authorized patient context from Epic, while a new plugin connects to nine public sources including PubMed, ClinicalTrials.gov, DailyMed, and CMS Coverage.
The Epic piece is worth understanding a bit deeper. OpenAI appears to have built an app against Epic’s standardized and public APIs, with UCSF as the pilot partner. Health systems still need to deploy it one by one, approve permissions, configure access, and test it, so nothing super novel as some folks made it sound. For now it is also read-only, so ChatGPT can reason over the chart but cannot write back or take clinical actions. That makes clinical decision support an increasingly crowded battleground, with Epic, OpenEvidence, etc and the frontier labs all moving closer to the chart.
OpenAI also launched GPT-6 Astra, its most capable model yet. On HealthBench Professional, Astra scored 63.4%, ahead of GPT-5.6 Sol at 60.5%, Claude Fable 5 at 60.9%, and Gemini 3.8 Flash at 52.1%.
The pieces they seem focused on collecting are slowly starting to come together: frontier models, real medical data, and direct access to the clinical workflow. (link)
2/
OpenEvidence is going deeper into specialist AI
OpenEvidence is pushing further into oncology while also rolling out a new family of medical AI models.
Its oncology agent now combines NCCN treatment algorithms, ASCO guidance, and a precision oncology knowledge base from a leading cancer center, giving clinicians patient-specific genomic interpretation alongside the broader evidence base. More specialist agents are planned in genetics, cardiology, and neurology.
At the same time, OpenEvidence launched Osler, Sackett, and Snow, each built for a different level of depth. Osler handles fast point-of-care questions, Sackett digs further into the evidence, and Snow can spend 5+ minutes investigating harder cases. The names are a nod to medical history: William Osler helped shape modern bedside medicine, David Sackett pioneered evidence-based medicine, and John Snow helped found modern epidemiology.
Above them sits Darwin, which OpenEvidence calls its most intelligent model ever and the most advanced medical AI model in the world. In its published evals, Darwin topped the frontier-model baselines across all four benchmarks tested, including 100% on a MedQA set, 82.7% on HealthBench Professional, and 87.2% on NOHARM. Darwin, fittingly named after Charles Darwin, is still research-only while its safeguards are validated.
The direction is toward a digital multispecialty team, with different models and specialist agents working together rather than one general-purpose medical chatbot. (link)
3/
Oura gives us a consumer health pure play to watch
Oura’s IPO filing is a pretty impressive look at where the wearables market is heading.
The company now has 5M paid members, more than $1.4B in trailing revenue, and grew 74% year over year through June. Members wear the ring about 23 hours a day, retention is roughly 85%, and Oura has accumulated nearly 42B hours of longitudinal biometric data. The customer base is 72% women, 30% Gen Z or younger, and 37% have household incomes below $100K.
Wearables are becoming an increasingly popular continuous data layer for health, capturing sleep, heart rate, activity, stress, fertility, and other signals between doctor visits. Oura’s bet is that AI can turn that stream into increasingly personalized and predictive health guidance. It already has 1,200+ integrations and is expanding into health plans, employers, fertility, metabolic health, and chronic care (ex. Counsel Health partnership earlier this summer).
That makes Oura a useful public-market pure play for us to watch. The combination of wearables + AI could be one of the clearest paths toward more proactive, continuous, and engaged healthcare, rather than a system that mostly waits until something goes wrong. (link)
Tools & Partnerships 🔧
Latest on business, consumer, and clinical healthcare AI tools and partnerships…
TOOLS
OpenAI launches GPT-6 Astra with stronger healthcare performance: The new model set a high on HealthBench Professional, outperforming GPT-5.6 Sol on complex clinician tasks while adding stronger reasoning, computer use, and multi-step workflow capabilities. (link)
AI catches heart disease in two seconds: Trained on 10.6M ECGs, the model analyzes a routine heart trace in under two seconds, detecting heart failure in 81% of cases and valve disease in 90%. A 590-patient trial across six hospitals is now underway ahead of potential NHS deployment. (link)
Limbic becomes first AI mental health company in FDA TEMPO: Its Unpacked service delivers clinician-supervised CBT to Medicare patients over the phone using an AI voice agent, with outcomes tracked through the agency’s new real-world evidence pilot. (link)
Cadence becomes second participant in FDA’s TEMPO pilot: HypertensionOS uses AI-assisted software to support clinician-supervised medication management for Medicare patients with Stage 2 hypertension. (link)
Dyson launches $499 AI-powered CameraJet toothbrush: A built-in macro camera analyzes 28 images per second to identify gaps between teeth, triggering a precision mouthrinse jet within 100 milliseconds to target hard-to-reach areas. (link)
Google launches Gemini 3.8 Flash for agentic workflows: The model improves reasoning, coding, and multi-step task performance while maintaining Gemini 3.7 Flash pricing, alongside a separate Flash Cyber model focused on cyber defense. (link)
Meta releases Muse Spark 1.3 for long-horizon agents: The updated model improves coding, reasoning, tool use, and multi-step workflows while reducing the number of tool calls needed for complex agentic tasks. (link)
Anthropic’s Mythos 5.1 sets new protein-design record: The model achieved nearly 50% hit rates across 12 protein targets, compared with typical rates of 10–15%, with binding affinities reaching up to 10x those of top outside competition entries. (link)
DrFirst automates prescription prior authorization with AI: The tool uses existing patient data to fully answer 63% of PBM question sets without clinician edits, with a median submission time of just 49 seconds. (link)
Anthropic launches safeguards for enterprise AI deployments: Enterprise Frontier Safeguards detects cyberattacks and misuse while letting organizations keep AI activity data in their own cloud storage under their own encryption keys, with healthcare organizations contributing to its development. (link)
AI predicts radiology no-shows using social determinants: University Hospitals Cleveland tested a model on 10,000 patients that reached a 0.76 AUROC, with targeted outreach to the highest-risk patients projected to save up to $6.2M. (link)
TotalSegmentator predicts patient traits directly from medical scans: The open-source imaging tool can now estimate weight, height, age, and sex from CT and MR scans in under a minute on a standard CPU. (link)
Nourish embeds generative AI assistant into its patient app: The 24/7 Health Assistant provides personalized meal planning, lab insights, and appointment preparation between dietitian visits, extending support beyond scheduled care. (link)
OpenEvidence evidence base gets a systematic review: An npj Digital Medicine review of 11 studies found generally relevant, evidence-supported answers and fewer fabricated citations than general-purpose LLMs, while performance was weaker on more complex clinical scenarios. (link)
CliniCARE-Bench puts medical AI agents through tougher testing: The new benchmark evaluates agents on longitudinal EHR investigations and finds defect-free performance can fall substantially below headline accuracy once errors across multi-step workflows are considered. (link)
Sutter expands ambient AI onto nursing floors: The health system is moving ambient documentation beyond outpatient physicians into inpatient nursing workflows after broader deployment across primary care and specialty settings. (link)
Healthcare’s AI ROI measurement gap keeps widening: Deloitte finds health systems are rapidly increasing AI investment, but relatively few can consistently measure and attribute concrete financial returns to those deployments. (link)
Nearly 90% of biomedical papers show signs of AI assistance: An analysis of recent PubMed articles found linguistic indicators consistent with widespread LLM use in scientific writing, underscoring how quickly generative AI is entering biomedical research workflows. (link)
TRIMEDX adds AI agents for medical device cybersecurity: The new capabilities identify cybersecurity vulnerabilities, match them to affected medical devices, and prioritize remediation across healthcare organizations’ clinical equipment fleets. (link)
Corti launches sovereign AI infrastructure for Europe: The platform provides governed AI access through an OpenAI-compatible interface while emphasizing data sovereignty, zero retention, compliance controls, and enterprise cost management. (link)
MaineHealth defends Palantir tool amid nurse protests: The health system says it uses Palantir technology to identify and challenge insurance claim denials as nurses protested the deployment across multiple cities. (link)
Gyde launches AI to automate insurance renewals: Renewals AI handles needs assessments, client outreach, plan-change analysis, scheduling, and other workflows to help Medicare and ACA brokers retain existing members while continuing to grow new business. (link)
PARTNERSHIPS
Mayo Clinic + Rainfall Health: Mayo Clinic deployed Rainfall Health’s AI compliance platform to support care coordination and quality reporting under CMS’s Transforming Episode Accountability Model. (link)
Merge Labs + Butterfly Network: Merge Labs partnered with Butterfly Network to use its ultrasound-on-chip technology in developing less-invasive, ultrasound-based brain-computer interfaces. (link)
HealthEx + Eight Sleep: HealthEx partnered with Eight Sleep to let members connect their clinical records with Pod biometrics, creating a more complete view of sleep and health data. (link)
Novant Health + Doximity: Novant Health partnered with Doximity to integrate Doximity Ask’s evidence-based clinical AI directly into Epic, giving clinicians secure answers at the point of care. (link)
Healwell AI + Converge Health Iowa: Healwell AI partnered with Converge Health Iowa to deploy its AI-powered SMART Search and SMART Summary tools across the state’s health information exchange, helping clinicians surface relevant information from longitudinal patient records. (link)
Franciscan Health + Brown University Health + Microsoft: Franciscan Health and Brown University Health expanded their use of Microsoft’s agentic AI and Azure tools to improve clinical workflows, infrastructure, and AI governance. (link)
Komodo Health + NCQA: Komodo Health partnered with NCQA to use real-world patient data and AI analytics to accelerate the development, testing, and refinement of HEDIS quality measures. (link)
FDB + Tebra: FDB deployed its Script Agent with Tebra to turn ambient patient-visit conversations into structured prescriptions for clinician review, extending ambient AI from documentation into prescribing. (link)
Deal Desk 💰
Spotlight on latest capital raises, M&A, and investments…
FUNDING
Scan.com, a medical imaging network, raised $90M in equity funding. Noteus Partners led, joined by Aviva, Concord Health Partners, YZR Capital, and Oxford Capital. It also secured $130M of debt. (link)
Konko AI, a NYC and Costa Rica-based AI platform that helps healthcare providers manage patient communication and admin tasks, raised $6M in funding. Hi Ventures led the round. (link)
Norbert Health, whose physical AI control system turns robotic hardware into autonomous nursing assistants, raised $14M in Series A funding led by William A. Marino with Cardinal Group. (link)
Konko AI, an AI patient coordination platform for healthcare providers, raised $6M in funding led by Hi Ventures, with participation from LifeX Ventures, SquareOne Capital, GroundUp Ventures and Phoenix Fund. (link)
M&A
NVIDIA + Hugging Face: NVIDIA agreed to acquire Hugging Face for $12.93B, adding the open-source AI platform used by more than 18M developers and 200,000 organizations. (link)
Vitality + Icario: Vitality acquired Icario to create an end-to-end member engagement platform combining AI-powered personalization with prevention, care gap closure, and long-term behavior change across 19M members. (link)
LeanTaaS + Aidin: LeanTaaS acquired Aidin to connect its AI-powered hospital capacity platform with a post-acute referral network spanning 200+ hospitals and 21,000+ providers. (link)
Cureety + Reimagine Care: Cureety acquired Reimagine Care to combine precision oncology monitoring with conversational AI and virtual clinical care, creating a platform supporting 250+ cancer centers and 100,000+ patients. (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👇
