Clinical & Diagnostics
6 Health Systems Enhancing Care Delivery with Ambient AI Scribes
Intermountain Health reported a 27% reduction in time-in-notes per appointment using Dragon Copilot across clinicians with 10+ encounters between April 2024 and December 2025. A separate multicenter study found physicians using ambient AI scribes saw burnout rates drop from 51.9% to 38.8% after just 30 days. The systems analyzed used ambient listening models integrated directly into Epic EHR via the App Orchard, capturing and structuring clinical conversations in real time without requiring physician manual input.
IKS Health Debuts First-of-Its-Kind Agentic AI Platform at AMGA Annual Conference
MyCareHub uses an agentic architecture that self-orchestrates multi-step patient engagement workflows without requiring human-in-the-loop for each interaction. It integrates with Epic via the Epic Connection Hub using bidirectional FHIR-based data exchange. IKS describes it as the first production-grade autonomous patient engagement system combining care coordination, navigation, and adherence support in a single agentic layer — a meaningful technical step beyond rule-based patient outreach automation.
AI-Driven Nurse Staffing Can Cut Costs and Maintain Patient Access, Columbia Business School Study Finds
AI-driven predictive staffing models reduced hourly staffing costs by more than $160 per hour in an ED setting, annualizing to ~$1.4M per department. The models incorporate real-time patient acuity data, historical census patterns, and seasonal demand signals to generate hourly staffing recommendations. The researchers found that AI tools optimizing existing staff yield dramatically stronger ROI than substitution-focused approaches — a distinction they call critical for health systems navigating nursing union concerns about AI displacement.
How Amazon Connect Health Brings Agentic AI to the Point of Care
Amazon Connect Health uses agentic AI agents across voice and digital channels for scheduling, prior authorization, documentation, and patient routing. The platform integrates with major EHRs via FHIR APIs and runs on Amazon Bedrock for model infrastructure. AWS positions the ROI argument around AMA-documented data: physicians spend 1.5–2 hours on EHR documentation per clinical hour. The system supports multi-turn agentic task completion, enabling a single agent session to handle end-to-end workflows like prior auth submission and follow-up without human handoff.
Research & Science
Clinical AI Has Boomed — Stanford-Harvard State of Clinical AI Report Shows What Holds Up in Practice
The ARISE network report found multi-agent diagnostic frameworks achieved accuracy gains of 7% to over 60% compared to single-agent baselines across diverse clinical tasks. The evidentiary gap is the report's most striking finding: the vast majority of AI devices entered via device-modification pathways using existing safety evidence rather than new randomized trials, with only 2.4% supported by RCT data. The authors flag this as a systemic risk as clinical AI scales — making post-market surveillance the critical safety mechanism in the absence of pre-market clinical evidence requirements.
Merck and Mayo Clinic Announce AI-Enabled Drug Discovery and Precision Medicine Collaboration
The collaboration will train AI and ML models on Mayo's multimodal clinical datasets spanning genomic, proteomic, imaging, and EHR data to identify novel drug targets and accelerate candidate selection. The deal mirrors a January 2026 SOPHiA GENETICS-MD Anderson collaboration using AI-powered genomic analytics for oncology precision medicine. Pharma-health system data partnerships now account for over 30% of Q1 2026 healthcare AI deal activity, establishing the health system patient dataset as a strategic AI asset class.
Precision Oncology in the Age of AI: Lessons from AI-Driven Drug Discovery and Clinical Translation
The review covers AI-guided platforms integrating genomic, proteomic, and transcriptomic datasets through LIMS-connected pipelines to surface molecular disease mechanisms hidden from single-modality analysis. Generative AI for de novo molecule design and virtual screening are identified as the most transformative near-term tools; quantum computing simulations of protein-drug interactions are positioned as a 3–5 year horizon capability. The authors identify tumor heterogeneity, biomarker validation, and trial design for AI-selected subpopulations as the primary clinical translation bottlenecks — not model performance.
Deep Learning Integration of Pathology and Radiology in AI-Assisted Medical Imaging
The framework introduces the Adaptive Multi-Resolution Imaging Network (AMRI-Net) with an Explainable Domain-Adaptive Learning (EDAL) strategy for cross-modal feature alignment between radiology (CT, MRI) and pathology (H&E slide) data. The model operates across both modalities without requiring paired training samples for every case — a key advantage given the rarity of co-registered datasets in clinical practice. Performance reaches state-of-the-art on multiple cancer subtype classification benchmarks, with saliency-map explainability to support clinician trust and adoption.
Policy & Regulation
FDA "Cuts Red Tape" on AI-Enabled Devices and Wearables in Sweeping Oversight Overhaul
The guidance exempts Clinical Decision Support (CDS) software issuing single recommendations from device regulation, provided tools meet existing non-device CDS criteria including transparency about recommendation basis. Wearables tracking heart rate, blood pressure, and blood glucose for wellness purposes receive broader regulatory leeway. The framework aligns U.S. oversight with ISO 13485:2016 via QMSR updates. In 2025, FDA issued 295 new AI device authorizations (three in four imaging-related) — that pace is expected to accelerate materially under the new rules.
Aidoc Wins FDA Clearance for Foundation Model AI Covering 14 Acute Radiology Indications
CARE (Clinical AI for Radiology Engine) received clearance for 11 new indications combined with 3 previously cleared ones — all under a unified model architecture covering pulmonary embolism, intracranial hemorrhage, aortic dissection, pneumothorax, and others. The regulatory significance: rather than clearing each indication as a separate device submission, FDA's decision suggests a pathway where a single foundation model can be cleared across multiple downstream applications — potentially compressing multi-year multi-submission timelines to a single clearance event for broad-capability radiology AI platforms.
TEFCA Reaches Nearly 500 Million Health Records Exchanged as HHS Deploys AI to Reduce Burden
TEFCA uses FHIR-native architectures to enable standardized data sharing across Qualified Health Information Networks (QHINs). HHS's AI integration strategy targets administrative workflows — prior authorization, claims processing, and care gap identification — using AI models that operate on the live TEFCA data fabric rather than requiring separate warehouse pipelines. Pilot data indicates AI-assisted prior authorization reduces delays by an estimated 30–40% in participating health systems. The 500M milestone reflects rapid participation growth that now covers the majority of U.S. health systems.
Americans May Be Losing Trust in AI in Health Care, Survey Finds
The Ohio State University Wexner Medical Center survey found belief that AI makes healthcare more efficient also fell from 64% to 55% — despite growing operational evidence to the contrary. The trust erosion cuts across demographics and is not explained by age or technology literacy. Researchers attribute the decline to high-profile AI billing error stories, data privacy concerns, and patient invisibility into when and how AI is used in their care encounters. Findings add weight to policy calls for mandatory AI disclosure requirements analogous to existing informed consent frameworks.
Industry & Business
Digital Health Funding Hits $7.4B in Q1 2026 Driven by AI Drug Discovery and M&A Surge
Earendil Labs led the quarter at $787M — to scale a deep learning platform with 40+ therapeutic candidates. Other major rounds: Abridge ($300M Series E at $5B), Ambiance ($243M Series C at $1.04B), Function Health ($300M Series C at $2.2B), and Qualified Health ($125M Series B). AI's share of health tech funding has grown from 29% in 2022 to 55% in Q1 2026. Flagship-backed Generate:Biomedicines filed for Nasdaq IPO seeking up to $425M, signaling healthcare AI's transition from private capital to public markets.
UnitedHealth Group Is Making a $3 Billion Bet on AI — What Does It Mean for Patients?
The investment is distributed across Optum's analytics and technology divisions, deploying large language models to automate prior authorization decisions, streamline claims processing, and flag high-risk members for care management. The company targets the $500+ billion annual U.S. healthcare administrative cost burden. Critics including patient advocates and CMS officials have raised concerns that the same AI infrastructure built for efficiency could be tuned to increase denial rates — citing prior controversy over UnitedHealth's algorithmic prior auth practices under congressional investigation.
Jimini Health Raises $17M to Launch AI Mental Health Chatbot Sage for Complex Cases
Sage operates as a clinician-augmentation tool: it integrates into behavioral health organization workflows to surface AI-generated session summaries, risk flags, and between-session patient support for complex presentations including treatment-resistant depression, bipolar disorder, and co-occurring conditions. The architecture keeps all clinical recommendations under physician oversight — a direct response to Stanford HAI research showing autonomous AI chatbot therapy may lack efficacy and generate dangerous responses in high-risk populations. The $17M seed will fund two to three large BHO partnerships ahead of a Series A.
Social Buzz
25% of Americans Now Using AI for Health Info — 14 Million Skipped a Doctor Visit Because of It
The Gallup data arrives as OpenAI reports over 5% of all ChatGPT messages globally are healthcare-related. A viral patient behavior pattern has emerged: people uploading itemized bills to AI to identify duplicate charges and Medicare rule violations. Consumer AI health use operates almost entirely outside HIPAA — none of the top consumer AI tools (ChatGPT, Gemini, Perplexity) qualify as covered entities — creating a regulatory gap at the fastest-growing point of patient AI adoption. Healthcare IT leaders are flagging this as a governance risk with no current federal remedy.
AI in the Mental Health Workforce Is Met with Fear, Pushback — and Enthusiasm
NPR documents practices saving 10–15 hours weekly per clinician through AI administrative tools — routing referrals, predicting no-shows, flagging medication non-compliance — with zero patient-facing AI involvement. Simultaneously, Stanford HAI research showed AI therapy chatbots may contribute to harmful stigma and dangerous responses in high-risk populations. The story crystallizes the 2026 behavioral health AI reality: administrative AI is thriving in production; autonomous clinical AI remains largely undeployed due to evidence gaps and liability concerns. The workforce debate is likely to shape behavioral health AI regulation in 2026.
AI in Healthcare: Experts Sound the Alarm on Data Privacy and Patient Trust
The article identifies three interconnected issues: (1) patients often have no visibility into when AI is involved in their care encounter; (2) consumer health AI tools operate outside HIPAA protections, creating a regulatory gap at the fastest-growing point of patient AI contact; and (3) health systems are deploying "shadow AI" — tools adopted at departmental level without enterprise governance frameworks. Healthcare leaders are calling for mandatory AI disclosure requirements analogous to informed consent, and urging ONC to update information blocking rules to address AI-generated clinical recommendations as a distinct category of health information.