What You Should Know

• Global clinical AI platform Heidi raises $340M in total new capital, comprising a $100M Series C equity round at a $900M valuation and a $240M growth investment from General Catalyst’s Customer Value Fund (CVF).
• The Series C equity financing was led by long-term investor Blackbird, with continued participation from Phoenix Court (LocalGlobe, Latitude, and Solar funds), Point72 Private Investments, and Headline, bringing Heidi’s cumulative funding to $436.6M.
• The $240M CVF growth capital structure provides dedicated go-to-market and balance-sheet capacity to fuel enterprise expansion while allowing equity capital to focus directly on proprietary R&D, clinical safety infrastructure, and regulatory submissions.
• Expands beyond ambient transcription into supervised agentic workflows; the suite includes Heidi Evidence (context-aware point-of-care clinical research, answering over 10 million queries since March 2026), Remote (a proprietary clinical wearable microphone), and Dictate (voice-to-text).

Operational Traction, Scale, and Financial Velocity

Founded in Melbourne, Australia, by physician-technologist Dr. Thomas Kelly, Heidi has demonstrated rapid commercial expansion across global public and private health systems:

• Encounter Volume & Clinical Reach: The platform now supports 2.8 million patient visits per week across 190 countries and 110 languages. In aggregate, Heidi has supported more than 175 million patient visits (up from 73 million at its Series B) and 67 million clinical hours, more than tripling usage since September 2025.
• ARR Acceleration: Annual recurring revenue (ARR) grew fiftyfold from $1M to $50M by April 2026 within 24 months.
• Enterprise Adoption & Flagship Contracts: Heidi maintains an enterprise activation rate of approximately 62%. Key institutional accounts include:
  • United States: Massachusetts-based integrated delivery network Beth Israel Lahey Health.
  • United Kingdom: Selected as the sole supplier for NHS England Midlands—the largest clinical AI procurement in NHS history.
  • Australia & New Zealand: Enterprise deployments across The Royal Children’s Hospital Melbourne, Children’s Health Queensland, Metro South Health, and a nationwide implementation across every emergency department in New Zealand.

Platform Expansion: From Scribes to Supervised Action

Heidi is expanding its software and hardware product surface beyond the visit consultation to capture the full clinical workday:

• In-House Model Architecture: Unlike competitors that wrap commercial frontier models via third-party APIs, Heidi runs the vast majority of its transcription, medical entity extraction, and clinical note generation on proprietary in-house models. This architecture allows the platform to optimize compute latency, lower inferencing costs, and tailor models specifically to clinical tasks.
• Heidi Evidence (Point-of-Care Intelligence): Launched in March 2026, Evidence surfaces context-aware clinical practice guidelines and medical literature at the point of care, having answered over 10 million clinical queries to date.
• Hardware & Modality Diversification: Introduced Remote, a purpose-built wearable audio-capture device for sterile surgical environments and ambulatory exam rooms, alongside Dictate, an integrated voice-to-text dictation tool.
• Supervised Agentic Workflows: New capital will accelerate autonomous agents that navigate multi-step clinical tasks—such as prepopulating order sets, referral drafting, and pre-encounter record synthesis—with the supervising clinician retaining final review and sign-off. (Note: These upcoming agentic capabilities will initially exclude the UK and EU markets due to regional regulatory pathways).

Enterprise Compliance and Safety Standards

To support deployments across enterprise delivery systems and public health services, Heidi complies with global regulatory and data privacy frameworks, including HIPAA (U.S.), GDPR (EU), NHS information governance standards (UK), and the Australian Privacy Principles. The platform holds ISO 27001, SOC 2 Type II, Cyber Essentials Plus, and the newly established ISO 42001 certification for Artificial Intelligence Management Systems.

“From the day I started Heidi, the ambition was always bigger than writing doctor’s notes. I imagined that AI would sit alongside clinicians and complete real work under their supervision,” said Dr. Thomas Kelly, Co-founder and CEO of Heidi. “That’s the shift we’re now making: from documenting care to helping clinicians act on it. This round is about getting an AI Care Partner to every doctor in the world, and putting clinicians back at the center of care with their patients.”

What You Should Know

• The Private Equity Stakeholder Project (PESP) released its comprehensive report, the 2026 State Healthcare Policy Review: Tracking Private Equity Oversight and Reform, examining nationwide legislative efforts to regulate private equity control across provider networks.
• As of August 2026, six states had enacted nine distinct healthcare oversight laws—one each in Washington, Vermont, and Delaware, and two each in Illinois, Maine, and Connecticut—while California and Oregon began implementing sweeping statutory frameworks passed in 2025.
• Highlights private equity’s footprint across acute and ambulatory delivery, tracking 1,029 private equity-backed healthcare transactions in 2025, with PE firms owning more than 550 hospitals (accounting for nearly 1 in 8 private, non-government U.S. hospitals) and operating over 500 healthcare facilities through joint ventures with non-profit health systems.
• Identifies a legislative shift: state policy has advanced beyond basic hospital change-of-ownership notice to target indirect levers of control, including management services organizations (MSOs), debt-financed dividend recapitalizations, real estate investment trust (REIT) sale-leasebacks, and parent holding company structures.

Four Key State Policy Approaches in 2026

The report groups emerging state legislative and regulatory strategies across four primary mechanisms:

1. Transparency, Ownership Disclosure, and Reporting

States expanded pre-closing notice requirements to unmask upstream holding structures and MSO agreements before deals become irreversible:

• California (AB 1415): Implemented in 2026, bringing PE groups, hedge funds, MSOs, and provider holding entities under Office of Health Care Affordability (OHCA) review. Draft regulations require notice when an investor takes a 5% or greater stake in debt, equity, or liabilities, or gains operational veto rights.
• Washington (HB 2548): Broadened transaction notice to capture outside changes in majority ownership/control, substantial asset transfers, and hospital sale-leasebacks, granting the Attorney General expanded investigative review windows.
• Illinois (HB 5000 & HB 4728): HB 5000 makes transaction notice permanent and covers upstream entities owning or controlling two or more Illinois healthcare providers, including out-of-state entities generating $10M+ from Illinois patients. HB 4728 mandates quarterly ownership, debt, staffing, and fee disclosures for developmental disability service providers owned by asset managers.
• Rhode Island: Implemented an Attorney General rule requiring 60 days’ notice for deals creating medical practices of eight or more providers or involving PE-backed MSO acquisitions.
• Maine (LD 2202): Mandates that healthcare entities submitting federal Hart-Scott-Rodino (HSR) antitrust filings provide concurrent filings to the state Attorney General.
• Connecticut (SB 196 & SB 125) & Vermont (H.583 / Act 133): Enacted recurring ownership disclosures, MSO organizational filings, and mandatory governance attestations.

2. Administrative Approval and Enforcement Authority

While several states sought the power to block or condition deals, statutory authority remains rare:

• Maine (LD 2201): Enacted the year’s only new administrative review and approval framework specifically targeting private equity, hedge fund, or qualifying MSO transactions. Requires 180 days’ advance notice and empowers the Department of Health and Human Services (DHHS) to approve, condition, or block transactions, mandating comprehensive reviews for deals exceeding $100 million in assets.
• Pending & Stalled Efforts: Pennsylvania’s review bill (HB 1460) was narrowed in the Senate (raising asset transfer thresholds from $10M to $25M), while Hawaii’s public-interest review bill (SB 3175) failed to advance. New Jersey companion bills (S4216 / A5204) targeting REIT hospital leases remain pending in committee.

3. Targeted Financial and Real Estate Prohibitions

To prevent capital extraction practices highlighted by high-profile provider bankruptcies (e.g., Prospect Medical Holdings in Connecticut and Rhode Island, and Crozer Health in Pennsylvania):

• Connecticut (SB 196): Prohibits acute care hospitals from entering into sale-leasebacks involving their main campus or inpatient real estate.
• Delaware (SB 313): Enacted a temporary ban (through July 1, 2028) on for-profit entities acquiring control of nonprofit acute care hospitals, while permanently subjecting hospital real estate sales and encumbrances to the state’s Healthcare Conversion Act.
• Failed Restraints: Maine rejected limits on hospital debt-to-equity ratios above 50% (LD 2198) and REIT sale-leaseback bans (LD 2197), while Rhode Island held a bill (S2950) requiring private equity buyers to post an upfront one-year operating expense bond.

4. Modernizing the Corporate Practice of Medicine (CPOM)

States are updating CPOM doctrines to prevent non-physician investors from using MSOs, “friendly PC” structures, and administrative services agreements to dictate clinical operations:

• California (SB 351 Enforcement): Entered 2026 enforcing restrictions barring PE firms and hedge funds from interfering with clinical judgment or operational autonomy. The Attorney General secured major enforcement restructurings and penalties against corporate entities, including settlements with Aspen Dental (2Mpenalty/300k restitution) and Carbon Health.
• Oregon (SB 951): Prohibits MSO contracts from exercising de facto control over provider compensation, staffing, scheduling, billing, and payer contracting, facing its first court test in Lane County emergency department staffing.
• Vermont (H.583 / Act 133): Barred PE groups and hedge funds from interfering with clinical standards, diagnoses, treatment, working hours, and medical staff hiring/firing decisions.
• Failed CPOM Bills: Proposals establishing strict licensee-only practice ownership or anti-retaliation provisions failed to advance in Washington (SB 5387), Maine (LD 2199), and Rhode Island (S2459).

Strategic Takeaways for Dealmakers and Health Systems

The 2026 legislative cycle demonstrates that state legislatures are shifting toward regulating operational control rather than strictly direct equity ownership. While wholesale bans on private equity ownership largely failed to pass, states are successfully establishing pre-closing review gates, hospital real estate sale-leaseback restrictions, and corporate practice enforcement that limit traditional financial engineering tactics. Healthcare private equity sponsors, MSO aggregators, and joint-venture hospital operators face increasing compliance overhead, mandatory holding disclosures, and state-level antitrust review windows that lengthen deal closing timelines.

For information about the report, visit https://pestakeholder.org/reports/2026-state-healthcare-policy-review/

The forward deployed engineer model, embraced across enterprise technology, is built on a seductive assumption: that the gap between a technology’s potential and its adoption is primarily a technical gap. A missing integration. A misconfigured workflow. A feature not yet built. Fix the code, and the organization will follow.

In most industries, that assumption is partially right. In healthcare, it is almost entirely wrong. 

In part, healthcare organizations contribute to the problem because most evaluate technology vendors the same flawed way. They audit the roadmap, assess the integrations, and negotiate the contract. They rarely ask enough how the vendor will be embedded in their operations after go-live, and what discipline that ongoing partnership requires. 

That question turns out to matter far more than the feature set. The vendors who will define the next decade of healthcare automation are not the ones shipping the best product. They are the ones who have figured out how to make any product stick inside the complexity of how healthcare actually operates.

The Misdiagnosis

The truth about healthcare technology in 2026 is that the feature gap is real but secondary. EHRs, automation platforms, AI-powered front-office solutions: the core capabilities that could transform how physician groups and health systems operate are largely available today. What is not available, in most organizations, is the operational infrastructure required to make those capabilities real.

The actual gap is one of operational integration: misaligned incentives between clinical and administrative staff, fragmented accountability, change fatigue from failed implementations, and workflow debt (the accumulated weight of processes never redesigned, workarounds never documented, and handoffs never formally owned), so deeply embedded that no amount of software reconfiguration can surface it. A forward deployed engineer can reroute a call flow. They cannot get a front-desk coordinator and a medical assistant to agree on who owns a part of the new intake process. They can configure an automation trigger. They cannot ensure that the staff behavior required to activate it actually changes and stays changed.

Health Enterprise Partners’ 2026 Healthcare Executive Survey makes this structural gap visible. Three in four organizations report active GenAI programs, yet most remain layered onto existing processes rather than embedded into redesigned workflows. Across five organizational readiness dimensions, average scores cluster between 2.7 and 3.1 on a five-point scale. These are not technology scores. They are organizational scores, and they reveal exactly what forward deployed engineers are not equipped to fix.

As one Ochsner executive put it, implementing AI keeps surfacing conversations that are fundamentally about people and workflow, not technology. The question the industry isn’t asking loudly enough is whether it’s spending enough time rethinking how things are done, rather than just applying AI to make a bad process more efficient.

What Forward Deployed Operations Actually Look Like

A forward deployed operator sits at the intersection of process design, stakeholder alignment, and technology enablement. They speak the language of the clinic and the platform, but their primary tool is not code. It’s change management. They measure success in workflow adoption rates, staff behavior change, and sustained operational throughput, not feature deployment velocity.

The distinction matters in practice. A forward deployed engineer asks: what does the system need to do? A forward deployed operator asks: what does the organization need to become?

That is a harder question. It cannot be answered once at go-live and then considered closed. Operational transformation in healthcare is not a setup exercise. It is an ongoing one.

A Different Model in Practice

The organizations getting this right share a common structural approach. Rather than deploying engineers to configure software and move on, they embed small teams of forward deployed operators directly inside practices and physician groups, backed by a deeper bench of technologists and change management specialists supporting the work behind the scenes. The operator is not a consultant who hands off a playbook. They are a permanent fixture in the organization’s operating rhythm.

What that model reveals, over time, is how wrong the setup-and-handoff assumption really is. The operational change required to make automation work does not happen at go-live. It accrues. Workflows that seemed straightforward reveal hidden dependencies. Staff who were trained get promoted or leave. The coordinator who championed the new intake process moves on, and the adoption curve resets. None of this is a failure of the technology. All of it is a failure mode that only embedded operators, people who are in the organization rather than periodically visiting it, are positioned to catch and correct.

The implication is uncomfortable for much of the vendor community: technology and operational change cannot be sold separately and expected to produce durable results. The technology enables the outcome. The operator sustains it. Until those two things are treated as inseparable, the industry will keep generating impressive pilot metrics and disappointing at scale.

About Frederik Mueller

Frederik Mueller is the CEO and Co-Founder of Third Way Health, an AI-human hybrid operations partner for physician groups and MSOs.

What You Should Know

• Abridge, the AI-native clinician intelligence platform deployed across more than 300 U.S. health systems, announced it was selected through a distribution partner to provide ambient clinical AI under a landmark U.S. Department of Veterans Affairs (VA) enterprise contract.
• The multiple-award Indefinite Delivery, Indefinite Quantity (IDIQ) contract carries a total ceiling of $775.72 million over five years across all eligible competing vendors, allowing individual VA medical centers and regional Veterans Integrated Services Networks (VISNs) to procure task orders.
• The contract provides access across the largest integrated healthcare network in the United States, which delivers care to more than 9 million Veterans across 1,380 healthcare facilities, including 170 VA Medical Centers (VAMCs) and nearly 1,200 outpatient care sites.
• Abridge is already operational across more than 75 VA medical centers following a successful nationwide pilot, supporting thousands of clinicians in outpatient primary care, a dozen clinical subspecialties, and the VA’s virtual Clinical Resource Hubs.

Operating Across a Split EHR Environment: VistA/CPRS and the Federal EHR

A key technical differentiator enabling Abridge’s selection is its demonstrated operational footprint across both of the VA’s operating electronic health record systems during a competitively awarded pilot:

• Dual-System Interoperability: Abridge is fully operational on both the legacy VistA/CPRS architecture and the new Federal EHR (Oracle Health), allowing clinicians to maintain a unified ambient documentation workflow regardless of where their facility sits in the modernization roadmap.
• Pilot Footprint: Through the pilot phase, Abridge deployed across more than 75 VA Medical Centers, serving thousands of VA clinicians.
• Clinical Setting Breadth: The system is actively deployed across ambulatory primary care, twelve medical and surgical specialties, and the VA’s Clinical Resource Hubs (CRHs)—the regional virtual care networks providing telehealth coverage to rural and underserved veterans.
• Multilingual Capability: Validated to support ambient clinical documentation in more than 28 languages across diverse veteran demographics.

Federal Security Governance and Data Protections

Deploying clinical AI across the Department of Veterans Affairs requires compliance with rigorous federal cybersecurity and privacy baselines:

• VA Security Authorization: Abridge’s deployment operates under formal VA security authorization, processing veteran protected health information (PHI) within dedicated commercial cloud environments governed by strict access controls and least-privilege principles.
• Patient-Controlled Ambient Capture: Clinical conversations are captured only with the veteran’s explicit consent, processing conversational audio into structured Subjective, Objective, Assessment, and Plan (SOAP) notes that must be reviewed and approved by the clinician before committing directly to VistA or the Federal EHR.

Expanding Beyond Note Generation: Context-Aware Clinical Intelligence

Abridge is leveraging the enterprise deployment to position its platform as an active clinical decision support and care continuity layer:

• Eliminating “Story Retelling”: Veteran care frequently crosses VAMCs, community care network (CCN) clinics, and virtual touchpoints. Abridge standardizes clinical documentation so downstream providers receive consistent clinical context without forcing veterans to repeatedly restate medical histories.
• Linked Evidence Point-of-Care Support: Extends beyond basic ambient dictation to surface real-time, context-aware clinical insights within native workflows—synthesizing data from the ongoing patient encounter, the historical EHR chart, and peer-reviewed medical literature.
• Enterprise Scale: The VA contract expands an enterprise footprint that will see Abridge support more than 100 million patient-clinician conversations across more than 300 commercial and academic health systems nationwide this year.

“It is an honor to support the clinicians who care for our nation’s Veterans. We’ve worked to earn trust by meeting them where they are and preserving clinical context across providers, specialties, care settings, and the EHR migration,” said Dr. Shiv Rao, CEO and Co-Founder of Abridge. “As a cardiologist, I’ve had the privilege of treating Veterans at VA medical centers in Ann Arbor and Pittsburgh. We’ve built Abridge to meet the scale and complexity that an enterprise-wide deployment requires.”

What You Should Know

• Healthcare access infrastructure leader Zocdoc launches an enterprise platform expansion that syndicates provider availability beyond its native marketplace (Zocdoc.com and mobile app) into a distributed Care Access Network.
• The platform allows healthcare providers to maintain a single integration that makes their real-time calendar availability instantly discoverable and bookable across external partner channels, including Google Gemini, Amazon Health AI, Yelp, Healthgrades, Blue Shield of California, general search engines, and commercial insurance directories.
• Zocdoc introduced a zero-barrier pricing model: any healthcare provider can join the network for free, with zero booking fees when existing patients book or when patients search for a specific clinician by name; providers incur a one-time fee only when Zocdoc facilitates a new patient discovery and confirmed booking.
• The strategic pivot addresses search fragmentation revealed in a new national Zocdoc survey, which found that 69% of patients seeking a new doctor check multiple channels, more than 1 in 3 have switched providers due to scheduling friction, and 43% would switch to a practice that offers easier digital booking.

The Care Access Network: Embedded Real-Time Scheduling

Rather than requiring healthcare practices to manage individual business profiles, directory feeds, and calendar connections across disparate third-party websites, Zocdoc aggregates provider availability through a single integration:

• Multi-Channel Distribution: Providers connected to Zocdoc are made bookable across external partner platforms, including Amazon Health AI, Google’s Gemini, Yelp, Healthgrades, and commercial payer portals such as Blue Shield of California.
• EHR and PMS Synchronization: The network syncs directly with more than 175 electronic health record (EHR) and practice management systems (PMS), ensuring that open calendar slots reflect real-time schedule changes and booked visits flow directly into clinical templates without front-desk rekeying.
• Rapid Appointment Access: While national average wait times to see an outpatient specialist exceed 30 days, Zocdoc connects patients to care with typical appointments occurring within 24 to 72 hours of booking.

Commercial Model: Zero-Fee Existing Bookings

Zocdoc paired the network expansion with a revised commercial pricing structure designed to accelerate provider network density:

• Free Network Participation: Any licensed medical provider can join Zocdoc’s scheduling directory at zero upfront cost.
• Elimination of Existing Patient Fees: Practices incur no booking fees when patients search directly for a specific clinician by name or book follow-up visits as an existing patient through Zocdoc channels.
• Pay-for-Performance Acquisition: Providers pay a one-time transaction fee only when Zocdoc facilitates net-new patient discovery and successfully completes an appointment booking.

Expanding Front-Office “Practice Solutions” and AI Phone Agents

Alongside external syndication, Zocdoc is deploying front-office software to help independent practices and medical groups convert organic website traffic and inbound telephone calls:

• Practice Solutions Suite: Provides white-labeled digital intake, automated insurance eligibility verification, digital patient communication threads, and branded website scheduling widgets (which have demonstrated a 25% to 30% lift in organic practice website bookings).
• Zo (AI Phone Assistant): Deploys an automated conversational voice agent to answer practice phone lines 24/7, reading real-time calendar availability and scheduling appointments without requiring front-desk staff intervention—countering call center abandonment where one in three patients report hanging up if placed on hold for even a minute.

“Keeping up with every new place patients search for care is becoming an impossible task for providers. With Zocdoc, now they don’t have to,” said Richard Fine, Chief Business Officer at Zocdoc. “We take on that complexity so practices and health systems can spend less time managing channels and more time doing what they actually set out to do: care for patients.”

The “Action Layer” for Generative AI and Search

The expansion of the Care Access Network addresses a central bottleneck in modern healthcare consumer navigation: the gap between discovery and execution. While conversational LLMs and search engines excel at answering symptom queries, recommending specialties, and indexing local clinics, they lack native, two-way read/write access to clinical scheduling engines. By providing developer APIs, machine-readable availability feeds, and Model Context Protocol (MCP) server hooks directly to platforms like Amazon Health AI and Gemini, Zocdoc establishes itself as the transactional utility connecting AI front-ends with operational provider calendars.

Learn more and join here: https://www.zocdoc.com/business

What You Should Know

• Patient access and navigation leader DexCare, Inc. announced the acquisition of Mila Health, an AI-driven care coordination company whose conversational agents autonomously call, text, and chat with patients to manage scheduling, pre-visit preparation, and post-discharge follow-ups.
• The transaction represents DexCare’s second corporate acquisition, building on its 2022 purchase of Womp, Inc., and expands its capital-backed platform that has raised $146 million to date (including a $75M Series C led by ICONIQ Growth in 2023).
• Mila Health’s conversational AI layer runs directly on DexCare’s access data model—validated across more than 10 million completed patient bookings—which unifies electronic health records (EHRs), scheduling rules, subspecialty requirements, and institutional clinician preferences into a single standardized rulebook.
• DexCare’s footprint encompasses 57 million covered patients across all 50 states, serving major integrated health systems including Kaiser Permanente, Piedmont, Texas Health Resources, and Tampa General Hospital, while helping health systems schedule 40% more appointments with existing resources and reduce time-to-care by five days.

Pairing Governed Access Rules with Autonomous Conversational Agents

Spun out of Providence Health in 2021, Seattle-based DexCare has raised $146M in capital (including a $75 million Series C led by ICONIQ Growth in 2023) and serves 57 million patients across health system clients such as Kaiser Permanente, Piedmont, Texas Health Resources, and Tampa General Hospital. Financial terms were not disclosed. The transaction marks DexCare’s second acquisition, following its 2022 purchase of e-commerce search and booking engine Womp, Inc.

The acquisition merges DexCare’s foundational access infrastructure with Mila’s agentic conversational capabilities:

• Codified Scheduling Governance: DexCare’s platform standardizes complex provider scheduling rules, clinical templates, location rules, and referral constraints across more than 10 million bookings. At Tampa General Hospital, for instance, DexCare codified 260 distinct scheduling policies while identifying more than 50 operational gaps.
• Traceable, Multi-Channel Outreach: Built as an API-first framework, Mila deploys multimodal agents across voice phone calls, SMS, and web chat. The agents handle overdue visit recall, verify subspecialty eligibility, coordinate booking times, walk patients through pre-procedure prep, and execute post-discharge check-ins.
• Auditable Clinical Guardrails: Unlike generic foundation models that hallucinate or execute without an audit trail, Mila operates strictly within the health system’s pre-approved access rulebook, providing traceable decision logic for why an agent booked a specific slot, escalated a call, or navigated a patient to an alternate site of care.

Demonstrated Operational and Financial Performance

Across live deployments with payers, providers, and software platforms, Mila has generated significant operational gains over traditional call centers:

• 54% Outbound Response Rate: Achieves a 54% patient response rate on outbound outreach campaigns—an 80% improvement compared to standard manual call center benchmarks.
• 50% Reduction in No-Shows: Automated reminders, multi-channel prep instructions, and interactive rescheduling cut clinic no-show rates in half.
• 18% Net Revenue Expansion: Increases health system top-line capture by accelerating patient throughput, converting overdue care gaps into booked visits, and maximizing operating room and procedural slot utilization.

Eliminating the “Agentic AI Context Gap”

The acquisition underscores an evolving consensus in healthcare AI: autonomous agents cannot function reliably without deep, institutional systems of record. While commoditized voice AI platforms can hold natural-sounding conversations, they typically fail in health systems because they lack real-time visibility into complex provider scheduling rules, credentialing boundaries, insurance nuances, and EHR templates.

“Mila gives our data a voice, and our data gives Mila direction,” said Matt Blosl, CEO of DexCare. “AI solutions are everywhere, but health systems quickly realize that building the infrastructure, scaling the pilot, and managing the agent’s boundaries is untenable. Most agents lack contextual awareness to connect a health system, to know the relationships between service lines and the nuances and preferences of staff, locations and providers. Any agent can sound incredibly human, but ours understands the health system it speaks for.”

By anchoring Mila’s conversational agents directly inside DexCare’s governed access data model, DexCare provides its agents with the institutional context needed to navigate patients without manual human intervention—moving healthcare access from fragmented point solutions toward a unified, automated system of action.

What You Should Know

• Healthcare and education technology provider Ascend Learning announced the acquisition of M7 Health, an AI-powered clinical scheduling and workforce management platform utilized across large academic medical centers, community facilities, and rural hospitals nationwide.
• The transaction creates an integrated enterprise workforce platform connecting education, credentialing, competency management, frontline leadership engagement, and dynamic shift deployment under a single enterprise umbrella.
• M7 Health’s algorithmic platform auto-balances hospital shift rosters, forecasts clinical staffing demand, and dynamically recruits internal clinicians to fill open shifts, driving documented operational outcomes including a 60%+ reduction in administrative scheduling burden, up to 40% lower premium labor spend, and a 30% reduction in nurse turnover.
• M7 Health integrates alongside Ascend Learning’s specialized enterprise portfolio brands, including StaffGarden (competency, professional governance, and clinical ladder tracking) and Laudio (frontline leadership workflow automation, digital rounding, and tiered huddles).

Bridging Clinical Education with Daily Shift Deployment

Led by CEO Dr. Lissy Hu (founder of CarePort Health and former President of Connected Networks at WellSky), Ascend Learning is assembling an end-to-end clinical workforce suite that connects education, credentialing, management, and staffing:

• Pre-Licensure and Credentialing Base: Ascend touches over 60% of U.S. nursing schools (via ATI Nursing Education) and certifies more than 245,000 allied health professionals annually (via the National Healthcareer Association).
• Competency and Progression Tracking: Integrates with StaffGarden, Ascend’s platform for tracking nursing clinical ladders, skills competencies, and ongoing credentials.
• Frontline Leadership Engagement: Aligns with Laudio, Ascend’s leadership platform that automates digital rounding, recognition, and tiered huddle workflows for nurse managers.
• Daily Intelligent Scheduling Layer: Incorporates M7 Health to translate competency profiles and career progression milestones directly into daily shift assignments, closed-loop gap filling, and automated schedule balancing.

M7 Health Platform Capabilities and Operational Benchmarks

Founded by CEO Ilana Borkenstein, MBA, RN, M7 Health was built by frontline nurses to replace static scheduling spreadsheets and legacy workforce management tools:

• Predictive Staffing & Balancing: Leverages AI models to forecast unit-level patient demand, auto-balance schedules, and match open shifts to available nurses based on clinical skills, fatigue rules, and personal scheduling preferences.
• Quantified Labor and Operational Metrics: Across academic medical centers, community facilities, and rural hospitals, M7 has demonstrated a 60%+ reduction in administrative scheduling time, a 30% reduction in nurse turnover, and a 35% to 40% reduction in premium/travel labor spend, alongside staff fairness scores exceeding 94%.
• Rapid Deployment: Complements existing hospital HRIS, payroll, and timekeeping systems (such as Kronos/UKG and Workday), delivering measurable operational ROI within 90 days of go-live.

Strategic Rationale: Unifying Healthcare Talent Supply and Demand

Healthcare staffing vendors have historically treated shift scheduling as an isolated logistics puzzle, optimizing purely for coverage without visibility into nurse burnout, competency development, or career aspirations.

By rolling M7 Health into Ascend’s broader ecosystem, Ascend creates an integrated continuum spanning talent creation (nursing education and licensure prep), talent qualification (StaffGarden), talent engagement (Laudio), and talent deployment (M7). This longitudinal visibility enables hospitals to schedule nurses based not just on who is available, but on who has the appropriate clinical certifications, whose career goals align with specialized unit training, and which shift patterns mitigate bedside turnover.

What You Should Know

• A commissioned Total Economic Impact (TEI) study conducted by Forrester Consulting in August 2026 revealed that Experian Health’s Patient Access Curator protected $50.4M in revenue over three years for a modeled composite health system.
• Over a three-year implementation window, the platform reduced coordination of benefits (COB) denials by 40%, eligibility denials by 35%, and registration-related denials by 20%.
• The financial analysis evaluated a composite integrated delivery network modeled from five customer interviews, representing a U.S. health system generating $5 billion in annual revenue, employing 20,000 staff, and serving 700,000 patients annually.
• By Year 3, deployment of the AI-powered coverage intelligence engine produced an 80% reduction in time spent on manual insurance discovery, releasing approximately 10,400 hours annually (equivalent to five full-time employees and $887,000 in redirected workforce capacity across three years).

Front-Door Denial Reduction and Direct Cash Acceleration

By automating patient intake data curation—validating demographics, active coverage, coordination of benefits (COB), and Medicare Beneficiary Identifier (MBI) records before encounters take place—the platform delivered significant front-end denial reductions over the three-year modeled period:

• Coordination of Benefits (COB) Denials: Reduced by 40%, mitigating disputes over primary vs. secondary payer order.
• Eligibility Denials: Decreased by 35%, intercepting inactive policies and terminated plan enrollments at registration.
• Registration-Related Denials: Dropped by 20%, correcting mismatched demographic markers, misspelled names, and miskeyed policy IDs.
• Cash Flow Acceleration: Yielded $82.2M in accelerated cash collections and generated a 5% reduction in days in accounts receivable (A/R days), alongside a 30% increase in upfront self-pay collections by Year 3.

Labor Reallocation and Operational Efficiencies

Beyond top-line revenue defense, the study quantified significant administrative labor savings by eliminating manual insurance searches and reducing reliance on outsourced BPO clearinghouses:

• 80% Cut in Insurance Discovery Time: Freed up approximately 10,400 hours annually across a typical 25-person patient access team—translating to the capacity of five full-time employees (FTEs) and roughly $887,000 in redirected labor value over three years.
• 45% Reduction in Outsourced Denial Management Spend: Yielded an estimated $2.25M in external vendor savings for health systems with a baseline $5M annual BPO budget.
• 10% Front-End Staff Productivity Lift: Automated real-time verification relieved front-desk registrars from manual phone calls and portal hopping, cutting downstream billing rework.
• Risk-Adjusted Present Value: Translated into $11.5M in net present-value (NPV) benefits over three years, which the report contextualized as equivalent to funding annual wages for 115 registered nurses, conducting 115,000 primary care visits, or purchasing seven new MRI systems.

Shifting RCM Left

The findings demonstrate a structural transition across revenue cycle leadership: shifting investment away from reactive, post-bill denial recovery toward proactive pre-service clearance. While legacy RCM strategies deployed downstream accounts receivable teams to chase denied claims through payer appeals, health systems are increasingly using AI-driven verification engines at intake to eliminate preventable data errors at the source. This upstream automation protects operating margins, insulates health systems against payer audit clawbacks, and spares patients from surprise out-of-network balance bills caused by misassigned insurance.

What You Should Know

• Global strategic advisory firm PwC published a market analysis titled “Unlocking the Women’s Health Opportunity: A New Era in Fertility and Reproductive Health,” projecting the global fertility sector to expand from $25 billion to $35 billion today to reach $35 billion – $45 billion by 2030, growing at a 6% to 8% compound annual growth rate.
• The market is experiencing a structural pivot from reactive infertility intervention to proactive, lifelong reproductive health management, exemplified by planned elective egg freezing cycles in the U.S. growing nearly 4x between 2014 and 2021.
• Demographic shifts are lengthening family-building timelines, with the average age of first-time U.S. mothers rising from 24.9 years in 2000 to 27.5 years in 2023, while family-building pathways expand via same-sex couples (more than doubling in treatment between 2012 and 2022 in the UK), single parents by choice, and donor-assisted conception.
• Investment dynamics have shifted from an initial wave of private equity clinic roll-ups—which deployed over $14 billion between 2020 and 2025—toward venture capital syndicates backing nearly 400 funding rounds in precision diagnostics, AI laboratory tools, and digital benefits platforms.

Key Market Dynamics Across the Five Core Sectors

• Provider & Care Delivery: Multisite private equity platform rollups have largely established the operational clinic infrastructure, but models are actively shifting to accommodate proactive, non-infertility patients. Elective egg freezing cycles in the U.S. nearly quadrupled between 2014 and 2021, prompting clinics to build specialized programs for healthy women in their 20s and 30s alongside virtual/hybrid navigation and integrated behavioral health support.
• Device & Diagnostics (AI in the Lab): AI models are entering embryology to replace subjective manual visual grading. In cited baseline research, embryologists averaged an implantation prediction accuracy of 51.9% (barely above chance), whereas deep learning time-lapse models reached 62.5% accuracy. Adoption of Preimplantation Genetic Testing (PGT) is expanding from high-risk cases into general IVF and elective preservation cohorts.
• Pharmaceuticals: Beyond mature gonadotropin protocols for ovarian stimulation, innovation centers on personalized stimulation dosing to curb hyperstimulation risk, next-generation follicle-stimulating hormone (FSH) formulations, and therapeutics targeting unaddressed conditions like Polyendocrine Metabolic Ovarian Syndrome (PMOS/PCOS, affecting 10% to 13% of women globally without an FDA-approved fertility drug) and recurrent pregnancy loss.
• Payer & Employer Benefits: Large enterprise employers increasingly treat fertility coverage as a core talent retention asset, with 25% of U.S. employers (500+ employees) offering or planning dedicated fertility benefits in 2026. At the policy level, 25 states and Washington, D.C. have enacted private insurance fertility coverage mandates as of March 2026, driving adoption across fully insured small- and mid-market commercial plans.
• Consumer Health: Direct-to-consumer (DTC) at-home hormone diagnostic panels, connected wearables, and cycle trackers are engaging women earlier. However, a sharp commercial divide is emerging between FDA-cleared, HSA/FSA-eligible diagnostic devices (such as quantitative progesterone confirmation) and unvalidated consumer wellness gadgets.

The Broadening Investment Thesis

Fertility remains among the most capitalized verticals in women’s health, attracting over $14 billion deployed between 2020 and 2025. While the previous cycle was dominated by private equity sponsor acquisitions consolidating fragmented independent practices into regional clinic platforms, modern capital flows are pivoting toward venture and growth equity. Backed by nearly 400 VC funding rounds over the five-year period, investment is concentrating upstream on AI-driven laboratory workflows, precision biomarkers, tech-enabled enterprise benefit managers, and condition-specific therapeutics.

Healthcare executives are being asked to approve AI investments through dashboards that emphasize activity: messages drafted, calls summarized, records reviewed and minutes saved. Those numbers show that a system is being used. They do not show whether the work reached a useful conclusion.

That distinction is easy to miss because healthcare workflows are divided into small steps. A patient message can be drafted while the request remains unanswered. A clinical note can be generated but still require substantial review before it is safe to sign. A scheduling request can be routed correctly without the patient receiving an appointment. The task may be complete even though the outcome is not.

Healthcare AI ROI should be measured by completed work, not by the number of tasks the technology touches. Activity matters, but it should not be mistaken for value.

The Gap Between Activity and Outcome

Healthcare work rarely follows a clean path. Information arrives late, cases fall outside standard rules and one decision often depends on several others. The harder cases reveal whether the technology has improved the operation or merely handled its easiest portion.

In patient communication, generating a reply may save time, but the organization gains little if the message does not resolve the question. In documentation, a first draft is useful only when it is accurate and usable by the clinician. In care coordination, placing a case in the right queue is not the same as making sure the next step occurs.

The same issue appears in administrative and financial work. A response may arrive faster while leaving important details unresolved, or a case may be categorized correctly while still requiring someone to decide what happens next. The unit of value is the workflow, not one action inside it.

Where ROI Becomes Overstated

The hidden cost usually appears after the dashboard stops counting. Staff review the output, compare it with another record, correct a field or reopen a case marked complete. Managers add checkpoints, and another department may absorb work that disappeared from the original queue.

If a tool removes five minutes from one task but adds review and follow-up elsewhere, the net gain is smaller than the headline figure. If one stage moves faster but the final resolution date does not change, the organization has improved motion rather than outcome.

The 2024 CAQH Index estimated that fully electronic administrative workflows could unlock about $20 billion in annual savings. Its recommendations also make an important distinction: electronic transactions must be supported by stronger workflows, particularly in prior authorization. The evidence concerns administrative automation broadly, but the lesson applies to AI as well. A faster transaction creates limited value when the surrounding work remains unresolved.

Executives should ask where work goes after the AI acts. Does it disappear, move to another team or return as an exception? A business case that cannot answer those questions is probably overstating ROI.

Define “Done” Before Measuring Value

The strongest AI programs begin with an operational definition of completion. That definition should reflect the outcome the organization is responsible for delivering, not the last action performed by the technology.

For patient access, completion may mean that the patient has an appointment and understands the next step. For clinical documentation, it may mean that the note is accurate, signed and usable for care. For care management, it may mean that a handoff has been accepted and follow-up is scheduled. For financial operations, it may mean that an issue has been resolved and the account can move forward without being reopened.

Once “done” is clear, leaders can track first-pass completion, staff intervention, reopened work and the time from the first action to the final outcome. They can also see whether the improvement is felt by patients and employees. These measures may be less dramatic than a volume dashboard, but they are more useful when deciding whether to expand, redesign or stop an AI initiative.

Human Review Is Part of the Operating Model

Human oversight should not be treated as an embarrassing exception to automation. Some healthcare decisions involve incomplete information, unusual circumstances or consequences that justify professional judgment. The aim is not to remove people from every step. It is to use their attention where it matters most.

Review time therefore belongs in the ROI calculation. Leaders should know how often staff intervene, what causes the intervention and whether the same exceptions keep returning. A tool that needs careful review on nearly every case may still be useful, but it is functioning as an assistant rather than an autonomous workflow. Its value should be described honestly.

CMS’s internal guidance offers a useful example of how a large healthcare agency approaches responsible AI. It calls for human oversight before AI output informs business decisions, along with continued monitoring, documentation and accountability. The guidance applies to CMS-related work, but the operating lesson is relevant: oversight is part of the cost of using AI reliably.

A Scorecard Leaders Can Defend

A credible scorecard should begin with end-to-end completion and then show what was required to reach it. First-pass completion, exception volume, reopened work, human-review time and time to final resolution tell a more useful story than tasks processed alone.

The financial measure should follow the operational result. Depending on the workflow, that might mean lower labor per resolved case, fewer delayed appointments, less documentation rework or faster completion of a patient request. It may also reveal that a system creates value in one area while adding burden in another. Leadership needs that information to improve the design.

Before approving the next phase of an AI program, executive teams should require a baseline for the full workflow and a clear definition of completion. They should compare total labor, exceptions, reopened work and final resolution times before and after deployment. Without that comparison, an ROI claim is little more than an activity report.

The board-level question should be simple: did the work reach a dependable conclusion with less delay, less rework and better use of human attention? Task automation tells leaders what the system did. Completed work shows what the organization gained. Healthcare AI should be judged by the second.

Sources

• CAQH, 2024 CAQH Index Report: Key Takeaways

• Centers for Medicare & Medicaid Services, Guidance for Responsible Use of Artificial Intelligence (AI) at CMS

About Ramkumar P

Ramkumar P is the Founder & CEO of Rytsense Technologies, where he leads the development of Agentic AI and Intelligent Automation solutions for Healthcare Revenue Cycle Management. With deep expertise in AI product development and enterprise automation, he helps healthcare organizations transform labor-intensive administrative processes into autonomous, scalable systems. He is passionate about practical AI adoption that delivers real business outcomes rather than experimental technology.

What You Should Know

• Angle Health, an AI-native healthcare benefits platform designed for small and midsize businesses (SMBs), announced a $600M equity financing transaction at a $2.7 billion valuation.
• The financing was led by global growth investor Vitruvian Partners, with participation from new backer Town Hall Ventures and existing investors Blumberg Capital, Portage Ventures, PruVen Capital, and Y Combinator.
• Angle Health currently serves more than 5,000 employer groups with customizable group health plans available across 47 states, supporting small businesses with as few as two employees.

Capital Structure and Transaction Details

• Primary Capital Inflow: The $200M Series C injection goes directly to Angle Health’s balance sheet to scale national distribution, deepen provider integrations, and enhance its AI underwriting infrastructure.
• Secondary Liquidity: The $400M tender offer provides substantial cash returns to early institutional seed and Series A investors as well as employees, coming less than 10 months after the company’s Series B round.
• Growth and Valuation Multiple: The $2.7 billion post-money valuation marks a major milestone for tech-enabled health plans, reflecting the company’s transition from an insurtech disrupter to a scaled, profitable payer.

Underwriting Economics and Operational Profitability

Founded in 2021 by CEO Ty Wang and Anirban Gangopadhyay (both former engineers at Palantir Technologies), Angle Health was engineered to replace manual small-group actuarial review with algorithmic underwriting:

• Profitability Milestones: Angle Health has achieved four consecutive quarters of EBITDA and GAAP net income profitability, expanding top-line revenue by 120% year-over-year.
• Scale and Premium Volume: Now manages nearly $1 billion in annualized premium-equivalents, serving over 5,000 employers across 47 states.
• Real-Time Group Underwriting: Angle’s Benefit Builder and Quote-to-Card engines allow commercial brokers to submit basic group censuses and generate firm, underwritten proposals within minutes—rather than the weeks typical of traditional carriers. Groups can onboard and issue digital insurance cards to employees almost instantly.
• Taming Renewal Increases: Leverages automated clinical steering and curated outpatient partnerships (covering high-cost specialty pharmaceuticals, home infusions, ambulatory surgery centers, and imaging) to keep median year-over-year renewal increases between 5% and 7%, compared to the 18% SMB market average.

Avoiding the Payer Pitfalls of Insurtech 1.0

First-generation venture-backed health insurers stumbled by chasing unprofitable, high-churn individual ACA exchange populations or aggressively bidding on Medicare Advantage risk pools where medical loss ratios (MLR) eroded margins.

Angle Health pivoted entirely away from that playbook by targeting the underserved commercial small-group segment (employers with as few as two lives). By packaging full-stack plan administration, level-funded risk arrangements, integrated telehealth, and AI navigation into a single package, Angle eliminates fragmented point-solution vendor bloat for small employers while maintaining strict underwriting discipline.

Healthcare AI has reached an inflection point. AI is rapidly moving from isolated demonstrations to production systems embedded in clinical, operational, and research workflows. A new generation of AI applications is emerging to support nearly every aspect of healthcare delivery, operations, and research.

As organizations deploy these capabilities at scale, they are exposing an invisible semantic challenge that has quietly existed for decades. Clinical information is translated repeatedly as it moves through the healthcare system. Each translation can subtly change its meaning, even though we usually treat the result as if nothing has changed. Few are discussing this problem, and even fewer are solving it.

Maintaining semantic continuity requires preserving clinical meaning across many forms, not just one. As clinical information moves through the healthcare system, it is continually interpreted and expressed in new forms, from conversations to clinical documentation, documentation to codes, codes to data, and increasingly, data to AI. Every transition is a transformation, and every transformation creates another opportunity for meaning to be lost, altered, or misrepresented. As these transformations accumulate, AI is increasingly asked to infer what is no longer explicit, no longer preserved, or no longer traceable to its original clinical meaning.

This raises a fundamental question. Where should healthcare semantic understanding live? Should every AI model reconstruct it independently from increasingly transformed data, or should it exist as a shared semantic infrastructure that every model can rely on?

The prevailing view is that increasingly capable models will eventually learn healthcare semantics. More clinical data, larger context windows, fine-tuning, and agentic workflows will progressively reduce the problem until semantic understanding simply emerges from the models themselves.

I believe the opposite. Healthcare semantics should not emerge independently inside every model. It should be preserved explicitly as shared semantic infrastructure that is traceable, reusable, and governed independently of whichever LLM or agent happens to consume it. Every AI model should reason from the same trusted semantic foundation rather than reconstructing it independently.

Even if future foundation models perfectly understood clinical meaning and were fluent in healthcare semantics able to translate between all vocabularies and ontologies, they would still face an important practical limitation. Healthcare depends on hundreds of shared vocabularies and coding systems for procedures, diagnosis, prescriptions and more. A model may correctly recognize methotrexate as a drug and return several associated National Drug Codes (NDCs). Yet herein lies the problem, methotrexate has more than six hundred NDC codes. Production systems for, say, patient finding or comparative effectiveness studies, often require every valid NDC, every descendant SNOMED concept, every applicable ICD code, or every relevant CPT code. That is not a reasoning problem. It is a fidelity and completeness problem. Healthcare AI requires both reasoning and completeness. LLMs increasingly provide the former. Semantic infrastructure must provide the latter.

This is not a limitation of AI. It is a separation of duty. Models reason over clinical meaning. Semantic infrastructure translates, preserves, and governs it. As foundation models continue to improve and agents become increasingly autonomous, that separation becomes more important, not less.

What does semantic infrastructure actually do?

Consider the seemingly straightforward task of identifying patients with idiopathic pulmonary fibrosis (IPF). To a clinician, the diagnosis may seem obvious. To an AI model, the clinical concepts are recognizable. Yet accurately identifying an IPF population across real-world data is anything but straightforward.

An IPF patient may be represented through pulmonology notes, radiology reports describing a usual interstitial pneumonia (UIP) pattern, pathology findings, pulmonary function tests, medications, ICD diagnosis codes, SNOMED concepts, and longitudinal patterns of care. Other patients may carry similar diagnoses but ultimately have connective tissue disease-associated interstitial lung disease, chronic hypersensitivity pneumonitis, or another fibrotic lung disease. No single representation tells the whole story.

A capable language model can recognize these concepts and reason about them. Production healthcare systems must first translate clinical meaning across disparate clinical representations and coding systems while preserving semantic fidelity and ensuring completeness.  Diagnoses become ICD or SNOMED codes. Medications become RxNorm or NDC identifiers. Laboratory observations become LOINC codes. Procedures become CPT or HCPCS codes. Every translation must be comprehensive, traceable, governed, and reproducible before AI can reliably reason and act upon them.

That is the role of semantic infrastructure. It establishes a trusted semantic foundation by resolving concepts across clinical language, coding systems, and data sources into a consistent, traceable understanding. It preserves semantic continuity as information changes form, supports deterministic retrieval of complete patient populations and concept sets, and provides every downstream model and agent with the same semantic foundation from which to reason.

Why does this matter now?

The stakes change dramatically once AI begins acting in healthcare workflows rather than simply answering questions in a chat interface. Historically, researchers and practitioners have compensated for gaps in clinical meaning. Clinicians recognize ambiguous documentation. Medical coders resolve inconsistencies. Analysts reconcile disparate coding systems. Real-world data scientists spend months translating, harmonizing, validating, and governing clinical concepts before they ever become evidence. Those activities are not incidental. They embody the necessary semantic work that makes healthcare data usable and trustworthy.

Agentic AI changes that equation. Increasingly, AI will not simply summarize information or answer questions. It will trigger workflows, recommend treatments, generate evidence, draft regulatory submissions, and coordinate decisions across healthcare systems. Every one of those actions depends on accurate and comprehensive semantic understanding of the underlying clinical information.

As AI assumes greater responsibility, the cost of semantic ambiguity changes. What was once an analyst’s inconvenience or a manual reconciliation becomes an operational dependency. The future of healthcare AI depends on two fundamentally different capabilities: 1) probabilistic reasoning and 2) deterministic semantic infrastructure. Confusing one for the other is becoming one of the industry’s biggest blind spots.

About Glynn Dennis

Glynn Dennis leads Kythera Labs Life Sciences and BioPharma initiatives, including research and product development. Glynn brings over two decades of expertise in Data & AI to Kythera Labs, having served in numerous scientific, data, and AI leadership roles across the industry, including at NIAID, Genentech, Bio-Rad Laboratories, and AstraZeneca.

What You Should Know

• The 2026 Healthcare Workforce Barometer, conducted online by The Harris Poll on behalf of Strategic Education, Inc. and its subsidiary Workforce Edge, surveyed 1,514 full-time direct patient care employees and 304 healthcare employers between June 12 and July 1, 2026.
• The study reveals that 59% of the total U.S. healthcare workforce is likely to look for a new role in the next year, while 70% of Gen Z healthcare workers plan to explore a new job over the same timeframe.
• These turnover intentions arrive as the Health Resources and Services Administration (HRSA) projects a national shortfall of nearly 500,000 critical healthcare workers by 2038, further compounded by the fact that one in three registered nurses is currently over the age of 50.

The Employer Perception Gap: Underestimating Workforce Flight

Hospital leadership consistently underestimates how actively clinical staff are preparing to walk out the door:

• Mobility Blind Spot: While 59% of clinicians report plans to look for new roles, employers estimate that only 39% of their workforce is actively looking.
• External Departure Risk: 42% of healthcare workers plan to look for positions outside their current organization, whereas employers believe only 32% are exploring external exits.
• Generational Turnover Exposure: 46% of employers identify early-career/Gen Z workers as their most difficult cohort to retain. Gen Z turnover intent (70%) substantially outpaces older peers: 64% of Millennials, 49% of Gen X, and 31% of Baby Boomers.

Gen Z Mobility: Fast Advancement vs. Long-Term Commitment

Contrary to the executive narrative that early-career clinicians lack institutional loyalty, the data reveals an advancement-driven workforce seeking long-term stability:

• The Stability Paradox: Despite 70% eyeing near-term role changes, 95% of Gen Z healthcare workers rate job stability and long-term security as vital, and 65% expect to stay with a single employer for five or more years.
• Speed and Progression: 89% of Gen Z workers say clear opportunities for advancement are critical (versus 80% across all generations), and 71% state the speed of career progression matters directly to them.
• The Career Trust Deficit: 47% of employers admit that a lack of advancement, development, or education opportunities drives staff departures. Yet only 26% of employees trust their employer a great deal to support their long-term career growth, and just 23% view their employer as an active career partner.

Education Benefits as an Underleveraged Retention Lever

While tuition reimbursement and upskilling programs consistently demonstrate retention ROI, most health systems fail to operationalize workforce education effectively:

• High Employee Appetite: 75% of healthcare workers want to pursue degree programs, clinical certifications, or skill-based credentials, and 81% want to participate in employer-sponsored education programs.
• Demonstrated ROI vs. Strategic Neglect: 89% of employers offering education benefits report measurable improvements in retention and mobility. However, less than half (48%) maintain a comprehensive workforce education strategy, and only 52% treat workforce education as a core strategic talent investment.
• Awareness Disconnect: 77% of employers report offering education benefits and estimate 69% employee awareness; in reality, only 58% of frontline staff know these benefits exist.

The AI Cooling Trend: Declining Employer Urgency and Lingering Distrust

The 2026 barometer highlights a marked cooling in enterprise AI sentiment across healthcare operations:

• Declining Executive Urgency: In 2025, 89% of healthcare employers believed AI skills were critical for workers to remain competitive; in 2026, that figure dropped 10 percentage points to 79%. Similarly, employers agreeing they have a duty to equip staff with AI skills dropped 11 points to 78%.
• Frontline Skepticism: Only 48% of healthcare workers feel comfortable using AI-based tools, 39% report AI has had zero impact on their day-to-day workflow, and just 45% trust that AI tools will yield tangible benefits for patients.
• Perceived Job Insulation: 66% of healthcare workers believe they face lower risk of AI job replacement than workers in other industries—anchored in the belief that direct patient interaction and complex clinical judgment cannot be automated.

What It Means for Healthcare Executives

For health systems operating under tight margins, relying exclusively on sign-on bonuses, agency travel staffing, and passive tuition assistance portals is failing to stem early-career churn. Younger clinicians are not rejecting long-term hospital employment; they are rejecting static, slow-moving career tracks. As enterprise AI implementation faces an operational “trust gap,” health systems that structure transparent, accelerated clinical ladders and actively guided education pathways will be far better positioned to insulate their bedside care capacity against the looming 500,000-worker shortage.

The full survey can be found here: HealthcareWorkforceBarometer.com.

What You Should Know

• Arizona-based nonprofit health system HonorHealth (operating nine acute-care hospitals, an extensive medical group, and outpatient networks across Greater Phoenix) has expanded its enterprise partnership with automated care operations leader Qventus by joining the AI Solution Factory.
• The AI Solution Factory is a co-development engagement model that embeds Qventus transformation engineers on-site with health system frontline clinicians and IT teams to design, prioritize, and deploy custom artificial intelligence solutions.
• The expanded alliance addresses severe macroeconomic pressures where modern health systems operate on razor-thin operating margins of 1% to 2% alongside persistent clinical staffing shortages, provider burnout, and declining reimbursement rates.
• The co-developed architecture designs autonomous “AI teammates” capable of listening, reading, speaking, writing, understanding, and executing tasks across structured and unstructured clinical data directly within native electronic health record (EHR) workflows.

The AI Solution Factory Co-Development Architecture

Rather than deploying rigid, off-the-shelf software packages, the AI Solution Factory model embeds Qventus transformation personnel directly within provider operations:

• Embedded Operational Engineering: Qventus AI teams work on-site alongside frontline clinicians, case managers, and hospital informatics leaders to identify operational friction points and co-design custom automation routines.
• Autonomous “AI Teammates”: Solutions co-developed within the Factory function as multimodal agents capable of ingesting structured and unstructured data across the EHR—listening, reading clinical notes, evaluating operational signals, and autonomously executing administrative actions (such as automated discharge milestone tracking, surgical block scheduling, and perioperative authorization status checks).
• Speed to Quantified ROI: Focuses engineering exclusively on measurable economic levers: expanding acute bed capacity, increasing surgical suite contribution margin, reducing avoidable patient days, and trimming direct operational costs.

Health System Vendor Consolidation Around “Systems of Action”

The HonorHealth expansion highlights an enterprise IT trend: healthcare provider organizations are shifting IT spend away from static reporting dashboards toward unified “systems of action” that operate natively on top of existing EHR systems of record (such as Epic):

• Multi-Module Platform Stickiness: Across Qventus’s footprint of more than 150 hospital facilities, nearly half of all enterprise clients now deploy multiple Qventus modules (spanning inpatient flow, emergency department operations, and perioperative optimization).
• Long-Term Enterprise Alignment: The vendor reports that 100% of its client base considers Qventus an ongoing component of their core enterprise IT architecture.

Operationalizing Agentic Healthcare Workflows

While basic generative LLMs have commoditized draft generation, translating algorithmic reasoning into reliable clinical and administrative execution remains difficult. Standard healthcare workflows contain unique compliance and clinical edge cases that break off-the-shelf software.

By utilizing a co-development model anchored inside a 9-hospital regional health system, Qventus secures rapid iteration loops to harden autonomous agents, while HonorHealth captures specialized automation configured precisely to its regional operating constraints without absorbing the full R&D burden internally.

“Qventus provides technology solutions that support key workflows within HonorHealth’s technology environment,” said Kim Post, DNP, MBA, RN, EVP and COO at HonorHealth. “Every year, our benchmarks improve, our processes get more efficient, and we find new opportunities to optimize care operations because Qventus is continuously learning alongside us. Each solution we’ve deployed has reinforced our confidence to go deeper, and the AI Solution Factory gives us the ability to take that partnership further — co-developing solutions that address our specific challenges in ways no off-the-shelf tool ever could.”

For years, healthcare leaders have asked if artificial intelligence can predict what will happen next, such as who might be readmitted, which patients could get worse, where resources are needed, and which interventions could help. Today, we can answer many of these questions, but a tougher question is what a healthcare organization should actually do with these predictions.

This is where much of healthcare AI still falls short.

The industry has invested heavily in electronic health records, cloud platforms, data warehouses, analytics, and machine learning. However, better data and more advanced models do not always lead to better decisions. The next step for healthcare AI should be making accurate predictions and focusing on turning those predictions into timely, measurable, and responsible actions.

A model that performs well does not always create real value.

A predictive model might work well during development but still fail to deliver value in real-world use.

A 2024 review in JAMA Network Open looked at 43 machine-learning algorithms used in primary care. The researchers found little public evidence about how these AI tools were implemented or how well they met quality standards. Only 12 out of 43 algorithms reached about half of the maximum evidence score in the review.

This difference is important.

Healthcare leaders are not interested in just a high score on a technical chart. They want to know if a prediction leads to better decisions and if those decisions improve outcomes, reduce unnecessary use, improve access, or make operations more efficient.

For example, consider a readmission-risk model. Flagging a patient as high risk is only just the first step. Someone must choose the right intervention, decide if it fits, deliver it on time, and check if it made a difference.

Without this decision step, a prediction is just another number on a dashboard.

What’s missing is decision intelligence.

Healthcare organizations should view AI as a tool for making decisions, not just for making predictions.

A good decision-intelligence system brings together five key parts:

Prediction > prioritization > intervention > measurement > learning.

Prediction spots risks or opportunities.

Prioritization decides which signals matter most, based on clinical importance, available resources, and potential impact.

Intervention turns insights into specific actions.

Measurement checks if the action made a real difference.

Learning completes the cycle by sending results back into the organization’s analytics and decision-making processes.

This approach also changes how we measure AI’s success.

Instead of just asking, “How accurate is the model?” healthcare leaders should also ask:

• Did the prediction change behavior?
• Did the resulting intervention improve an outcome?
• Did the intervention create unintended consequences?
• Was the model useful within the existing workflow?
• Did performance remain reliable after deployment?
• Did the system create value relative to its cost?

These questions are as much about business and operations as they are about technology.

Putting these ideas into practice is the next big step.

Recent studies highlight the gap between promising AI models and their long-term use in real settings.

A 2026 review in npj Digital Medicine looked at real-world uses of deep-learning systems in healthcare. The review found that research on how these systems are implemented is still limited. Most studies checked clinical outcomes, adoption, and appropriateness, but only one looked at costs, and none studied long-term sustainability.

This should change how organizations judge healthcare AI.

An AI project should not end once a model passes validation. Instead, deployment should be an ongoing process that includes workflow design, human review, monitoring, measuring outcomes, and managing the model.

This matters because healthcare is always changing. Patient populations shift, clinical practices evolve, payment incentives change, and the way data is collected also changes.

A model that works well today might not work the same way tomorrow.

Data quality still matters, but it is just the starting point.

This does not mean data quality is unimportant. It is essential.

However, focusing only on getting more data can hide a bigger problem: figuring out which data should guide which decisions.

Having more data does not always lead to better decisions. Healthcare organizations need data that is relevant, timely, easy to understand, and tied to a clear decision.

The World Health Organization’s recent work on AI in health highlights a similar idea. AI should support human judgment, not replace it. Implementation should include transparency, oversight, cross-disciplinary teamwork, and risk-based governance.

This principle should apply not just to clinical AI, but to all health services.

Moving from dashboards to decisions

The next step for healthcare analytics should focus on supporting decisions, not just creating dashboards.

A hospital aiming to reduce avoidable readmissions does not need another list of high-risk patients. It needs a system that helps answer which patients should get an intervention, what kind of intervention, when it should happen, who is responsible, and whether it worked.

A health plan looking to improve value-based care does not just need more predictive scores. It needs to turn those scores into clear priorities, assign resources, and track real changes in outcomes.

An executive deciding how to use limited resources needs more than a forecast. They need a decision framework that links the forecast to capacity, cost, risk, and expected results.

This is how healthcare AI can grow from a set of models into a true organizational strength.

The next competitive edge in the industry may not go to the group with the fanciest algorithm or the biggest data pool.

It may go to the organization that can reliably turn predictions into the right decisions, the right actions, and real results.

Healthcare is not lacking in predictions.

It has a chance to get much better at acting on them.

About Sachin Girdhar

Sachin Girdhar is an experienced healthcare analytics leader with over 15 years of experience turning complex healthcare data into practical plans that improve patient care and organizational performance. He specializes in predicting trends, Medicare Star Ratings, value-based care, health plan results, and using data to improve healthcare.

Sources

1. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2823631 
2. https://www.nature.com/articles/s41746-026-02358-2
3. https://www.who.int/publications/m/item/artificial-intelligence-for-health