Nolla Health: will use intake plus patient face photos to evaluate mild-to-moderate acne and recommend low-risk medications, with the goal of eventual autonomous prescribing. The service costs $4.99/month cash, avoiding insurance entirely. (link)

August AI: is testing prescription renewals across 100+ drugs, starting with lower-risk medications before moving into higher-risk ones. (link)

Expect Fitness: will test AI-guided pelvic floor physical therapy, with the longer-term goal of assigning plans without clinician review for low-risk patients. (link)

University of Utah Health + Intermountain Health: received umbrella agreements that create faster review pathways for future AI pilots, though each use case will still be vetted individually.

The three-year forecast became present tense. In 2025, 84% of executives expected GenAI to transform clinical treatment decisions within 3 to 5 years. One year later, 71% say it already has. Pasted text

ROI arrived 2x faster than expected. Buyers underwrote roughly 24-month paybacks but are seeing them in about 12 months, averaging 3.5x ROI. Around 40% of use cases beat expectations, and 54% show material ROI inside the first year. Pasted text

The back office is leading. Revenue cycle management posts 4.0x ROI, with 67% of solutions running as semi- or fully autonomous agents. Clinical use cases sit at 2.9x ROI, with just 4% operating at that level of autonomy. Pasted text

The workforce impact is starting to show. Half of organizations have already reduced headcount because of AI or plan to within six months, with cuts averaging 8% to 13% of affected functions. Pasted text

Buyers are consolidating. 42% have consolidated or are consolidating AI vendors, 60% have chosen a primary foundation model provider, and fewer than half of internally built tools are still maintained and in use. Development is shifting toward healthcare-specific AI vendors and AI labs. Pasted text

Clinical AI still has a trust problem. Only 46% trust AI-generated tools for clinical decision support, and 77% override suggestions more than half the time. Liability and reimbursement remain major blockers, especially for fully autonomous care. Pasted text

Heidi puts agents to work: Heidi II adds a computer-using agent that turns visit information into completed administrative tasks inside existing healthcare software. (link)

America.gov brings AI to government services: The chatbot draws on information from 29,000 government websites, with plans for actions such as Medicare enrollment in 2027. (link)

AKASA launches autonomous coding: Its new platform covers inpatient coding and clinical documentation, with outpatient facility support planned next. (link)

Tempus earns another FDA clearance: ECG-MR analyzes standard ECGs to flag signs of potentially undiagnosed mitral regurgitation, helping identify patients who may need further evaluation. (link)

Elsevier builds AI for nurses: ClinicalKey Nursing AI provides evidence-based answers to nursing questions, with traceable sources supporting its responses. (link)

Glass Health adds customization: Clinicians can shape responses through preferences, workflow skills, and preferred references, adapting the assistant to how they practice. (link)

HealthEdge unifies payer workflows: Its new AI-powered platform connects health plans’ core systems and operational workflows, targeting measurable financial outcomes. (link)

Trilliant launches an AI strategist: Oria combines language models with national all-payer claims data to answer healthcare strategy questions, plan analyses, and check its work. (link)

Welldoc connects care between visits: Its AI-powered care loop uses small language models and 500M+ health data points to personalize guidance and surface updates for care teams. (link)

Duke develops patient digital twins: The initiative combines imaging, AI, and wearable data, including a pilot aimed at predicting heart failure deterioration before symptoms or hospitalization. (link)

ARPA-H accelerates clinical trials: SURPASS combines predictive models, adaptive designs, and agentic trial operations, while STACK uses AI to speed site activation and expand research capacity. (link)

ChronicleBio targets chronic disease: Fidji Simo’s startup is combining AI with nearly 200TB of biological data from 1,000+ patients to identify disease subtypes and potential treatments for conditions including POTS and long COVID. (link)

Phylo maps drug targets: TargetAtlas assesses thousands of target-disease relationships, including work with the Michael J. Fox Foundation to prioritize potential Parkinson’s disease targets. (link)

Topos Bio tackles difficult proteins: Topos-2 models the movement of disordered proteins, while Topos-Bind predicts their interactions with potential drugs. (link)

Google watermarks AI-designed proteins: DeepMind says SynthID Bio embeds identifiers that remain detectable after physical synthesis while preserving protein function. (link)

Epic adds recovery comparisons: MyChart uses Cosmos data to help patients compare their recovery with others who underwent similar procedures. (link)

Kin Health adds monthly Recaps: The feature summarizes recent visits and outstanding tasks, helping patients keep track of their care. (link)

UT commits $25M to healthcare AI: REALHealthAI will support development, governance, and deployment across UT hospitals and medical schools, with attention to privacy and patient safety. (link)

OpenAI introduces always-on agents: Dots can handle ongoing work in the background; access in Healthcare workspaces is an admin-enabled beta that is off by default. (link)

Anthropic releases Sonnet 5.5: The company reports speeds 30%+ faster than Sonnet 5, with some tasks costing up to 30% less. (link)

Gemini finds a hospital data exposure: Google says Gemini 4 Argon identified a serious patient-data exposure through Wiz’s Scan for Good program that earlier models had missed. (link)

UC San Diego Health + Harvard + Anthropic: The partners are developing an open-source system to check the safety of patient-facing AI recommendations after deployment through ARPA-H’s ADVOCATE program. (link)

AdventHealth + OpenEvidence: The health system is embedding OpenEvidence into EHR workflows, giving clinical teams access to medical research and guidelines at the point of care. (link)

ACU-Serve + Tennr: The partnership automates intake, eligibility checks, prior authorization, and referral coordination for post-acute care and durable medical equipment providers. (link)

Epic + Anthropic: Epic used Claude Mythos in a six-week cybersecurity sprint to identify and patch vulnerabilities, including potential exploits involving open-source AI agents. (link)

OpenEvidence + Sage: The partnership brings research from 300+ peer-reviewed journals into OpenEvidence, with citations linking clinicians directly to the original articles. (link)

GoTo + athenaOne: An AI scheduling integration lets patients book, reschedule, or cancel appointments around the clock, with updates inside the EHR and escalation to staff when needed. (link)

Mendaera + Butterfly Network: Butterfly’s Ultrasound-on-Chip technology is integrated into Mendaera’s FDA-cleared Focalist robotic system, bringing real-time imaging into the procedural platform. (link)

AbbVie + Valkai: The partners will apply AI to targeted clinical development workflows, aiming to generate insights faster and shorten drug development timelines. (link)

CSL + AWS: The collaboration brings AI and cloud computing into drug discovery and clinical development, including target identification, protocol drafting, and regulatory workflows. (link)

Healthcare Triangle + AI World: A new joint venture plans to bring AI-powered nursing robots to North America, with Healthcare Triangle holding a 70% stake. (link)

Ambient documentation keeps spreading: PHTI reports that 71% of surveyed health systems have deployed ambient documentation in some capacity, with 41% reporting enterprise deployment. (link)

Patients still want human oversight: A Wolters Kluwer survey found 89% want human validation of medical AI, while 74% worry about privacy and 69% about inaccurate information. (link)

Google examines trust in medical AI: An AMIE paper argues that clinical trust requires rigorous real-world evidence and direct human experience alongside benchmark performance. (link)

Better prompts, better medical image performance: The RankingPE preprint reports gains of up to 16.2 AUROC points over accuracy-based prompt evolution, with results depending on strong medical vision encoders. (link)

The case for clinician-AI scientists: A Lancet Digital Health viewpoint calls for physicians trained in both medicine and AI to connect model development, clinical evaluation, and implementation. (link)

Full-body scans face an evidence question: Growing consumer demand is renewing debate over early detection, false positives, downstream testing, and whether screening delivers meaningful health benefits. (link)

Pennsylvania advances healthcare AI rules: The House passed HB 1925, which would require AI disclosure, human involvement in clinical and coverage decisions, and safeguards for privacy and safety. (link)

FDA proposes robotic surgery guidance: The draft outlines testing, cybersecurity, and clinical evidence expectations for robotic surgical devices, including risks associated with AI errors. (link)

Senate advances healthcare cybersecurity legislation: The bill would strengthen coordination and defenses, with particular attention to rural and other providers with limited resources. (link)

AHA flags gaps beyond HIPAA: The association urged Congress to address privacy and cybersecurity risks involving AI vendors handling health information outside HIPAA’s coverage. (link)

FTC examines frontier AI lab consumer risks: The agency is investigating OpenAI and Anthropic over potential consumer harms and safety concerns. (link)

Kaiser gives workers a role in AI decisions: Its labor agreement gives union representatives half the seats on a task force reviewing AI investments. (link)

RFK Jr. weighs AI second opinions: The HHS secretary suggested that failing to consult AI could someday be considered malpractice, describing a possible future expectation for care. (link)

Devoted Health, a Medicare Advantage insurer and medical group, has raised nearly $1.2B in fresh funding. The financing includes $555M in Series G funds co-led by private equity firm Advent International and Temasek. (link)

Ortet, a frontier AI lab for health co-founded by ex-Genentech AI researchers, launched with a $500M commitment from Thoreau. (link)

EliseAI, a NYC-based AI company automating housing and healthcare operations, raised $350M in funding. a16z and Bessemer Venture Partners led the round and were joined by Ontario Teachers' Pension Plan, Sapphire Ventures, and Navitas Capital. (link)

ElevenLabs, a voice AI platform, doubled its valuation to $22B through a $300M employee tender offer led by Wellington Management and T. Rowe Price. (link)

Outmarket AI, a San Francisco-based AI platform automating back-office work for insurance brokers, raised $34.5M in Series B funding led by SignalFire, with Fika Ventures, Permanent Capital Ventures, TTV Capital, and Dash Fund also participating. (link)

Colossal Biosciences, a de-extinction and genetic engineering company, is reportedly in talks to raise new funding at a $25B to $26B valuation led by TWG Global. (link)

Parakeet Health, a healthcare AI company, raised $10M in Series A funding led by Canvas Ventures, with participation from Blank Space Ventures, StoryHouse Ventures, and HMC INQ. (link)

Aleph Surgical, which builds embodied foundation models for fully autonomous surgery, raised a $7.5M pre-seed round led by Andreessen Horowitz, with BoxGroup, Reveille VC, Constellation, and others participating. (link)

2026 Nobel Prize in Medicine for discoveries behind optogenetics (link)

a16z: ~30% of S&P 500 report AI impact, but only ~2% track it (link)

Healthcare workers turn against Palantir (link)

Implementation Director at Clarium, an AI hospital supply chain platform

N/A | Remote (link)

Senior Software Engineer at Hippocratic AI, an AI agent platform for care

N/A | Menlo Park, CA (link)

Senior Infrastructure Engineer at Adonis, an AI revenue cycle platform

$190K – $210K | NYC (link)