The Healthcare AI Landscape for Small to Mid-Sized Organizations

Executive Summary

Artificial intelligence is penetrating healthcare at an accelerating but deeply uneven pace. Large health systems are driving headline adoption numbers, while the small and mid-sized organizations that deliver most of the nation's ambulatory, community, long-term, and behavioral health care remain significantly behind. This report catalogs documented AI use cases across five organization types, ranks them by adoption prevalence, maps the vendor ecosystem serving each, and surfaces the structural barriers that keep the adoption gap wide.

Key findings:

• Ambient AI clinical documentation is the dominant use case across all org types where it has been deployed, with 60+ vendors competing and cumulative encounter volume on leading platforms exceeding 1 million clinical encounters.
• RCM/billing automation is the fastest-growing use case by percentage-point gain: hospital adoption rose 25 points in a single year (36% to 61%), and is ranked among the top physician practice AI investment priorities.
• EHR-embedded AI is the primary delivery mechanism, 80% of hospitals using predictive AI rely on EHR-developed models, making EHR vendor choice a major determinant of AI access for smaller organizations.
• The independent/safety-net gap is structural and widening: system-affiliated hospitals adopt predictive AI at 86% vs. 37% for independent hospitals. For FQHCs, pricing models, workforce capacity, and data infrastructure are the primary blockers, not a lack of interest.
• LTC/AL remains the most structurally underserved segment, with documented AI adoption lagging ambulatory care significantly; behavioral health adoption data is thinner but directionally similar.
• Governance is a major unresolved layer across all segments, identified as a critical enabler yet frequently absent or inadequate, especially in resource-constrained organizations.

Methodology

This report synthesizes findings from federal survey data (ONC/ASTP), national association research (AMA, AHA, MGMA, HIMSS), peer-reviewed literature (PubMed/PMC, JMIR, npj Digital Medicine), foundation reports (California Health Care Foundation, Commonwealth Fund, Digital Medicine Society), sector publications (McKnight's, Becker's Hospital Review), and academic research institutions (IC² Institute/UT Austin). Sources are prioritized by methodological rigor and recency (2024–2026). Use case rankings reflect frequency of citation across authoritative sources, combined with available quantitative adoption data.

Section 1: AI Use Cases Ranked by Adoption Prevalence

1. Ambient AI Clinical Documentation (AI Scribes)

Prevalence: Highest across all documented org types where deployed

Ambient AI documentation, sometimes called AI scribing, converts real-time clinician-patient conversations into structured clinical notes, reducing or eliminating post-visit documentation burden. It is the single most widely cited and rapidly adopted AI use case in healthcare as of 2024–2026.

Among physician practices, MGMA polling in 2024 indicated that ambient documentation ranked as the top AI priority for a majority of medical group leaders, ahead of every other use case. AMA survey data reported that physician use of AI for clinical documentation roughly doubled between 2023 and 2024 (cited at approximately 13% to 21%; readers should confirm the exact figure against the AMA's published survey). Becker's Hospital Review reported that leading ambient AI platforms have reached cumulative totals exceeding 1 million clinical encounters, and that a multicenter study found physician burnout among ambulatory users declined from roughly 52% to 39% within 30 days of adoption. These vendor- and study-reported figures should be read as directional rather than definitive.

In FQHCs, Neighborhood Healthcare, a California FQHC serving approximately 100,000 patients across 30 facilities, deployed Nabla's ambient AI platform to 200 clinicians. Of 34,000 AI-generated notes, 76% were rated 4/5 or higher in accuracy and only 2% required editing. One nurse practitioner reduced after-hours documentation from 1–2 hours daily to completing notes same-day while seeing 16–20 patients. The Commonwealth Fund (2026) documented additional ambient scribing deployments at community health centers in rural settings.

Key vendors: Nuance DAX (Microsoft), Nabla, Abridge, Suki, Ambience Healthcare, DeepScribe, Freed, Heidi Health. 60+ vendors are now competing in this space.

Org-type status:

• Physician practices: Leading adopter; MGMA data shows 43% of groups added or expanded AI in 2024, with documentation as the primary driver.
• FQHCs: Early but growing; Nabla, Abridge gaining traction due to multilingual support (Nabla supports 35 languages in under 5 seconds per note).
• Hospitals (small/mid): Adoption expanding from large systems down; primary EHR vendors (Epic, Oracle) are integrating ambient tools.
• LTC/AL: Early stage; workflow differences reduce fit with current ambient scribing models.
• Behavioral health: Growing; some platforms support mental health visit templates. Recording sensitivity for BH sessions is a noted governance consideration.

2. Revenue Cycle Management (RCM) and Billing Automation

Prevalence: Fastest-growing by documented percentage-point gain; #2 overall

AI-assisted RCM encompasses automated coding, claims scrubbing, denial prediction and management, prior authorization support, and payment integrity. This was the single largest year-over-year growth area in hospital AI adoption, rising 25 percentage points, from 36% to 61% of nonfederal acute care hospitals, in one year, per ONC/ASTP (2024).

For physician practices, MGMA polling placed RCM AI among the leading investment priorities, behind documentation tools. AMA survey data shows administrative burden reduction as the most commonly cited AI opportunity area for physicians (reported at 57%), encompassing both documentation and billing.

RCM AI is also penetrating FQHCs and safety-net providers, though at a slower pace. The appeal is direct: denied claims and administrative overhead disproportionately burden smaller organizations with thin margins and limited billing staff.

Key vendors: Waystar, Nuvei Health, Change Healthcare (UnitedHealth Group), Availity, nThrive, Greenway Health, Veradigm (formerly Allscripts), Experian Health, athenahealth (Oracle), eClinicalWorks, M*Modal.

Org-type status:

• Hospitals (small/mid): Highest documented adoption; 61% as of 2024.
• Physician practices: High priority; core RCM functionality now embedded in most PM/EHR platforms.
• FQHCs: Moderate; FQHC-specific billing complexity (UDS reporting, cost-based reimbursement, sliding-fee scale) creates additional demand.
• LTC/AL: McKnight's (2025) cites RCM AI as a priority; Medicare/Medicaid billing complexity drives demand.
• Behavioral health: High unmet need; BH billing complexity (session limits, prior auth requirements, different code sets) makes RCM AI particularly valuable.

3. Appointment Scheduling AI

Prevalence: Third most documented; second-fastest growing among hospitals

AI-assisted scheduling encompasses intelligent appointment slot matching, no-show prediction, wait-list management, and automated patient outreach for scheduling gaps. Hospital adoption rose 16 percentage points, from 51% to 67%, in 2024, per ONC/ASTP. For physician practices, scheduling AI is typically delivered through patient engagement platforms or EHR add-ons. For FQHCs, automated scheduling reduces administrative phone volume, critical for understaffed front-desk operations.

Key vendors: Luma Health, Relatient, NexHealth, Zocdoc (enterprise), Kyruus, Phreesia, eClinicalWorks' healow, Epic's MyChart scheduling, Oracle's scheduling modules.

Org-type status:

• Hospitals/physician practices: Broadly deployed; often EHR-embedded.
• FQHCs: Strong fit; high no-show rates and language barriers make intelligent outreach valuable.
• LTC/AL: Lower priority; scheduling is less complex in facility-based care.
• Behavioral health: Growing; session frequency management and no-show prediction have direct revenue impact.

4. EHR-Embedded Predictive Analytics

Prevalence: Fourth; dominant delivery mechanism for most hospital AI

EHR-embedded predictive analytics includes in-workflow risk scores, clinical alerts, and population health flags surfaced within the EHR. As of 2024, 80% of hospitals using predictive AI rely on EHR-developed models, making EHR-embedded AI the most common delivery pathway and a major determinant of what AI capabilities are accessible to any given organization.

Common sub-use-cases include: sepsis early warning, readmission risk scoring, deterioration alerting (NEWS2, MEWS scores), length-of-stay prediction, and high-risk outpatient identification. The ONC/ASTP data shows that identifying high-risk outpatients for follow-up grew 9 percentage points year-over-year.

Key vendors: Epic (Cognitive Computing platform, sepsis prediction, deterioration index), Oracle Health (clinical analytics), Wolters Kluwer (UpToDate + clinical decision support integration), Health Catalyst, Arcadia, Lightbeam Health.

Org-type status:

• Hospitals: Primary adoption channel; system-affiliated hospitals at 86% adoption.
• Physician practices: Available through EHR, but uptake depends on value-based care contracts.
• FQHCs: Access depends entirely on EHR vendor; those on Epic or eClinicalWorks have access; those on smaller platforms may not.
• LTC/AL: Growing; fall risk prediction and pressure injury prevention are primary use cases.
• Behavioral health: Risk stratification for suicide and self-harm represents the emerging application.

5. Clinical Decision Support (CDS)

Prevalence: Fifth; widely available but utilization varies significantly

Clinical decision support encompasses drug interaction checking, diagnostic support, order set recommendations, preventive care reminders, and evidence-based protocol alerts. CDS tools are embedded in every major EHR, making them the most ubiquitous AI-adjacent tool in healthcare, though they are often not counted as "AI" in adoption surveys.

HIMSS/Medscape (2024) found that 86% of respondents reported their organization already leverages AI in some capacity, with clinical decision support and EHR-integrated tools among the top areas expected to grow. AMA survey data found that only 16% of clinicians report active use of AI specifically for clinical decision support, suggesting a gap between organizational availability and individual utilization.

Key vendors: Epic (CDS Hooks, BestPractice Advisories), Oracle Health, Wolters Kluwer (UpToDate, Medi-Span), IBM Micromedex, Stanson Health (acquired by Premier), Zynx Health.

Org-type status:

• Hospitals/physician practices: Broadly available; alert fatigue is a documented challenge.
• FQHCs: PCMH incentives drive CDS adoption; chronic disease management alerts are high-value.
• LTC/AL: Medication reconciliation CDS is particularly high-value given polypharmacy risk.
• Behavioral health: Limited; most CDS tools are not tailored to psychiatric medication management.

6. Prior Authorization Automation

Prevalence: Sixth; high growth trajectory; increasingly viewed as agentic AI's first major foothold

AI-assisted prior authorization automates the extraction of clinical criteria, generates supporting documentation, and in emerging deployments submits PA requests directly without human intervention. Becker's Hospital Review (2026) identified this as one of the primary areas where ambient AI is expanding into "agentic workflows", AI that takes action rather than just documenting.

Prior auth burden is among the most frequently cited administrative pain points across all org types. The transition from documentation AI to agentic PA automation represents a significant evolution in the near-term.

Key vendors: Cohere Health, Availity, Waystar, Voluware, pVerify, Epic's Coverage Discovery.

Org-type status:

• Physician practices/hospitals: High demand; most PA tools currently require clinician review of AI output.
• FQHCs: Significant burden; PA requirements on Medicaid programs create disproportionate load.
• Behavioral health: Extreme burden; behavioral health has among the highest PA denial rates; high unmet need.
• LTC/AL: Moderate; Medicare Advantage PA requirements are growing.

7. Patient Communication and Engagement AI

Prevalence: Seventh; broad but shallow deployment

AI-powered patient communication includes automated appointment reminders, post-visit follow-up, chronic disease check-ins, patient portal message triage, and AI-drafted responses to patient messages. EHR-native patient messaging tools (Epic MyChart, Oracle Health) now include AI-drafted response suggestions, which represent the most widely deployed form of patient communication AI.

Key vendors: Luma Health, Relatient, Phreesia, Klara, Epic MyChart AI features, Welldoc, Healthie.

Org-type status:

• Physician practices: Moderate adoption; patient portal message burden is a documented driver.
• FQHCs: High-value fit; multilingual patient populations make AI-assisted communication (with appropriate safeguards) attractive.
• LTC/AL: Family communication tools and care plan updates are primary applications.
• Behavioral health: Crisis triage and between-session check-in tools exist but carry significant clinical risk governance requirements.

8. Diagnostic Imaging AI

Prevalence: Eighth overall; dominant in radiology-intensive settings; limited relevance to most small/mid-sized org types

AI-assisted diagnostic imaging includes automated anomaly detection, image segmentation, triage prioritization, and computer-aided detection (CAD) for radiology, pathology, ophthalmology, and dermatology. This is one of the most mature and FDA-cleared AI segments in healthcare, with over 1,000 FDA-authorized AI/ML-enabled medical devices as of 2025.

This use case is largely not applicable to FQHCs, behavioral health organizations, and most physician practices, which do not operate imaging departments. Small/mid-sized hospitals with imaging capabilities are the primary relevant segment.

Key vendors: Aidoc, Nuance PowerScribe (Microsoft), Viz.ai, GE Healthcare AI, Philips IntelliSite.

Org-type status:

• Hospitals (with imaging): Meaningful adoption; stroke/PE/pulmonary embolism AI is FDA-cleared and clinically validated.
• Physician practices (with imaging): Limited to ophthalmology (diabetic retinopathy AI is CMS-reimbursed) and dermatology.
• FQHCs/BH/LTC: Largely not applicable.

9. Sepsis and Early Warning / Deterioration Detection

Prevalence: Ninth; high clinical value; hospital-centric

Sepsis early warning systems use ML models to flag patients at elevated risk based on vital signs, lab values, and EHR data. Epic's sepsis prediction model is the most widely deployed. These tools have generated significant clinical outcomes data alongside controversy about false positive rates and alert fatigue.

Key vendors: Epic (Sepsis Prediction), EarlySense, Philips (IntelliVue Guardian), Capsule Technologies.

Org-type status:

• Hospitals: Moderate-to-high adoption; directly tied to EHR platform.
• LTC/AL: Nursing facility deterioration prediction (fall risk, pressure injury) is the LTC analog.
• FQHCs/BH: Primarily acute care use case; limited applicability.

10. Remote Patient Monitoring (RPM) with AI Augmentation

Prevalence: Tenth; growing rapidly post-2020; high relevance to FQHCs and primary care

RPM tools collect physiologic data (blood pressure, glucose, weight, oxygen saturation) between visits. AI augmentation adds anomaly detection, alert generation, and outreach triggering. CMS expanded RPM reimbursement in 2020; utilization has grown significantly since. Texas safety-net providers specifically cited RPM in underserved/rural areas as a perceived AI benefit (IC² Institute/UT Austin, 2025).

Key vendors: Cadence, Current Health (acquired by Best Buy Health), Biofourmis, iRhythm (cardiac monitoring), Dexcom (continuous glucose, AI-augmented), Validic.

Org-type status:

• Physician practices/FQHCs: High relevance; aligns with value-based care chronic disease management.
• LTC/AL: Growing; AI-augmented fall detection (wearables) and bed sensors are specific applications.
• Behavioral health: Emerging; digital phenotyping (smartphone sensor data for mood/activity patterns) is in research stage.

11. Social Determinants of Health (SDOH) Risk Stratification

Prevalence: Eleventh; underreported but high growth trajectory; highest relevance to FQHCs

AI-assisted SDOH screening and risk stratification identifies patients at elevated social risk (food insecurity, housing instability, transportation barriers) and connects them to community resources. This is distinct from clinical risk stratification, it operates at the intersection of clinical and social data, and is most relevant to organizations serving high-need populations.

Research confirms that algorithm bias against underrepresented populations, driven by models that "did not consider social determinants of health to predict risk outcomes", is both a documented barrier and a driver for SDOH-aware AI development (JMIR Human Factors, 2024).

Key vendors: Unite Us, Findhelp (formerly Aunt Bertha), NowPow, Epic Social Care, Healthify, Pieces Technologies.

Org-type status:

• FQHCs: Highest relevance; HRSA UDS requirements and Section 330 accountability create demand.
• Community hospitals: Growing under CHNA requirements and value-based contracts.
• Behavioral health: High relevance; housing instability and food insecurity intersect heavily with behavioral health populations.
• LTC/AL: Transition planning and family resource support are emerging applications.

12. AI-Enabled Robotics and Physical Automation (LTC-Specific)

Prevalence: Twelfth overall; highest penetration in LTC/AL relative to other org types

A 2024 peer-reviewed scoping review in JMIR Aging catalogued AI-enabled robot adoption in long-term care, identifying applications in medication delivery, mobility assistance, social engagement (companion robots), infection control, and wayfinding. Documented barriers include technical complexity, ethical concerns about replacing human contact, resource limitations, and staff training gaps.

Key vendors: Aethon (TUG medication delivery robots), PARO (companion robot), Intuition Robotics (ElliQ for seniors).

Org-type status:

• LTC/AL: Primary and essentially sole deployment setting.

13. Behavioral Health Risk Prediction and Conversational AI

Prevalence: Thirteenth overall; rapidly evolving; significant governance requirements

Two distinct AI categories are active in the behavioral health space:

Risk prediction: ML algorithms trained on EHR data, behavioral patterns, and language patterns have demonstrated capacity to identify individuals at elevated risk for depression, psychosis, or suicidal ideation (2025 PLOS One; 2025 JMIR).

Conversational AI: Tools provide between-session support, CBT-based exercises, and symptom tracking. However, Social Current (May 2026) documented that AI chatbot use has been linked to exacerbating psychiatric symptoms, particularly among individuals prone to psychological dependency, including parasocial relationships, delusional thinking, emotional dysregulation, and social withdrawal. Organizations deploying conversational AI in behavioral health settings should have documented governance protocols in place before deployment.

Key vendors: Woebot Health, Wysa, Spring Health, Brightside Health, SilverCloud Health (acquired by Amwell, 2021), Lyssn (AI for therapy quality assurance).

Org-type status:

• Behavioral health organizations: Primary target; both risk prediction and conversational AI are segment-specific.
• FQHCs with integrated behavioral health: Growing application.
• Hospital ED/crisis settings: AI-assisted suicide risk screening at triage is an emerging application.

14. Workforce and Staff Scheduling Optimization

Prevalence: Fourteenth; documented in hospital and LTC/AL contexts

AI-assisted staff scheduling uses predictive modeling to anticipate patient census, match staffing levels, reduce overtime, and optimize shift assignments. McKnight's (2025) specifically identified workforce scheduling AI as one of the technology priorities for LTC providers in 2025 and beyond.

Key vendors: Shift.ai, ShiftMed, UKG (Ultimate Kronos Group), Infor (workforce management).

Org-type status:

• LTC/AL: High relevance; chronic staffing shortages and direct care worker turnover make AI scheduling optimization particularly valuable.
• Hospitals: Established; most use workforce management platforms with predictive components.
• Physician practices/FQHCs: Lower priority; provider scheduling complexity is generally lower.

15. Pharmacy and Medication Reconciliation AI

Prevalence: Fifteenth; widely available but under-counted in adoption surveys

AI-assisted pharmacy tools include automated medication reconciliation at care transitions, drug-drug and drug-allergy interaction checking, formulary compliance alerts, and prior authorization automation for specialty medications. These tools are often embedded in EHR and pharmacy management systems and are frequently not categorized as "AI" in adoption surveys, making this use case likely undercounted.

Key vendors: Medi-Span (Wolters Kluwer), IBM Micromedex, Epic MedRec, Surescripts (Medication History), DrFirst, Veradigm.

Org-type status:

• Hospitals: Broadly deployed via EHR integration.
• LTC/AL: High-value; polypharmacy in elderly populations makes medication reconciliation critical; transition-of-care medication errors are a primary quality metric.
• Physician practices/FQHCs: Available through EHR; adoption varies.

Section 2: Core AI-Enabled Platform Ecosystem

As the ONC/ASTP data confirms, 80% of hospitals using predictive AI rely on EHR-developed models. EHR platform choice is the single most consequential AI access decision for any small or mid-sized organization.

Section 3: Adoption Patterns by Organization Type

3.1 Hospitals (Small/Mid-Sized)

The ONC/ASTP 2024 data brief is the authoritative source. Among nonfederal acute care hospitals:

• 71% used EHR-integrated predictive AI in 2024 (up from 66% in 2023)
• System-affiliated hospitals: 86% adoption; Independent hospitals: 37% adoption (49-point gap)
• Critical Access Hospitals: 50% adoption vs. 80% non-CAH
• Small, rural, independent, government-owned, and critical access hospitals are the structural laggards

Top use cases by hospital adoption rate:

1. Scheduling optimization: 67%
2. Billing/RCM automation: 61%
3. Identifying high-risk outpatients for follow-up: growing rapidly
4. Inpatient deterioration detection
5. Sepsis prediction

The independent small/mid-sized hospital is most likely relying on EHR-embedded AI as its sole AI strategy, and most likely to be on a platform with weaker AI capabilities than Epic or Oracle.

3.2 Physician Practices

MGMA 2024 is the primary data source:

• 43% of medical groups added or expanded AI in 2024 (vs. 21% in 2023, a doubling)
• 66% of physicians used AI in 2024, up from 38% in 2023 (AMA)
• Top priorities (MGMA polling, approximate): ambient documentation leading, followed by RCM/billing AI and patient communication AI; specific percentages should be confirmed against the original MGMA materials
• Primary barriers: Training gaps, cost, clinician trust, EHR integration friction

The physician practice segment is bifurcating: larger multi-specialty groups and health-system-affiliated practices are adopting rapidly, while independent small practices and solo practitioners are moving much more slowly due to cost and implementation complexity.

3.3 FQHCs / Community Health Centers

FQHCs are the most underserved segment relative to their patient population needs. Key documented barriers (CHCF, 2025; IC² Institute, 2025):

• Pricing models don't work: Per-usage and per-visit pricing is prohibitive for organizations operating on cost-based reimbursement and thin grant margins
• Workforce gap: Most FQHCs lack data science staff or dedicated IT departments
• Infrastructure deficit: Data exchange capacity must be built before AI can be layered on
• Language and equity risk: FQHCs serve high proportions of non-English-speaking patients; AI tools trained primarily on English-language clinical data may perform differently across patient populations
• Liability uncertainty: Who bears financial risk for AI errors is unresolved

The Neighborhood Healthcare case (Nabla, 200 clinicians, 76% note accuracy, 35 languages supported) demonstrates that ambient AI is deployable and delivering value in the FQHC context when implementation is structured appropriately.

3.4 Long-Term Care / Assisted Living

LTC is the most structurally disadvantaged segment. Peer-reviewed analysis (PMC, 2025) indicates nursing and residential care firms adopting AI at low single-digit rates (reported in the range of roughly 3% to 4.5% across 2023–2025), well below ambulatory care. This figure is drawn from a single study and should be treated as directional. Key dynamics:

• EHR platform determines AI access: PointClickCare and MatrixCare dominate the LTC EHR market; their AI maturity lags the hospital EHR platforms
• Workforce crisis intensifies need: Chronic staffing shortages make AI-assisted scheduling and workload automation high-value
• Robotics has a foothold: AI-enabled companion robots, medication delivery robots, and mobility assistance tools represent a category unique to LTC
• RCM complexity: Medicare/Medicaid billing complexity, including MDS documentation requirements, creates demand for AI-assisted coding and compliance

McKnight's (2025) experts identified near-term technology priorities as staffing/scheduling AI, interoperability tools, RCM automation, and care delivery assistance.

3.5 Behavioral Health

Behavioral health is simultaneously one of the highest-need and highest-risk segments for AI deployment. Key dynamics:

• Documentation burden is severe: BH clinicians often maintain both structured diagnostic documentation and narrative therapy notes; ambient AI is well-suited but requires careful deployment to protect session confidentiality
• RCM complexity is extreme: BH billing is commonly reported among the specialties with the highest PA denial rates, shortest session limits, and most frequent payer disputes
• Conversational AI is advancing but requires governance: Social Current (May 2026) documented both promise and documented harm from AI chatbot use in vulnerable populations
• Risk prediction is maturing: ML-based suicide risk stratification is moving from research into clinical deployment but requires significant governance oversight
• The workforce crisis creates demand: Widely cited workforce projections point to significant behavioral health professional shortages through the late 2020s, increasing organizational interest in AI-augmented care models as a partial mitigation strategy

Section 4: Barriers to Adoption

The following barriers are documented across multiple sources and ranked by cross-study prevalence.

4.1 Trust (Most Cited Cross-Cutting Barrier)

JMIR Human Factors scoping review (2024) of 50 studies found trust as "the most critical element of AI adoption, it can either be facilitated or impacted by almost all the themes identified." Three dimensions:

• Black box problem: Clinicians require understanding of both the scientific and clinical basis of AI recommendations before adopting them
• Algorithm bias: Documented concern that models "did not consider social determinants of health to predict risk outcomes" and were "not representative of the patient population"
• Familiarity as a correlate: UT Austin/IC² data found that familiarity with AI tools directly correlates with higher trust; exposure reduces resistance

4.2 Financial Barriers

• Per-usage and per-visit pricing models don't translate to safety-net economics (CHCF, 2025)
• Inadequate funding was among the three primary barriers cited by Texas safety-net providers, alongside data privacy concerns and knowledge/training gaps (IC² Institute, 2025)
• Implementation costs (integration, training, workflow redesign) are not captured in licensing fees
• ROI timelines are too long for organizations facing quarterly cash flow pressure

4.3 Workforce and Technical Capacity

• 57% of safety-net providers expressed low or neutral confidence in their organization's capacity to integrate AI into workflows (IC² Institute, 2025)
• Most FQHCs, small physician practices, and BH organizations lack internal data scientists or AI-fluent IT staff
• Across the peer-reviewed AI adoption literature, limited staff capacity/training and workflow integration friction are among the most frequently cited barriers, appearing in a majority of the studies reviewed (JMIR scoping review, 2024)

4.4 Governance and Leadership Gaps

• Unclear leadership direction was among the most frequently cited organizational barriers across the peer-reviewed AI adoption literature (JMIR scoping review, 2024)
• Governance frameworks identified as a key facilitator, but the npj Digital Medicine (2026) systematic review of 35 frameworks found that virtually all existing frameworks skew toward large academic health systems and lack actionable pathways for resource-constrained organizations
• AHA (2025) specifically identified the structural governance gap between large systems and independent/CAH hospitals

4.5 Infrastructure and Data Quality

• FQHCs in underserved areas face poor broadband connectivity (CHCF documents specific California regions)
• Data fragmentation across multi-system regions prevents the longitudinal data access that AI models require
• Many smaller organizations are still addressing basic data exchange before AI layering is feasible

4.6 Regulatory Uncertainty

• Absence of clear regulatory frameworks cited across multiple studies as an adoption impediment
• FDA clearance pathway complexity for clinical AI tools creates organizational hesitance
• Liability for AI errors is unresolved, who is responsible when an AI-assisted coding error generates a compliance exposure?
• Patient acceptance is lower than providers expect: only 45% of safety-net providers believe patients would accept AI tools; 30% believe patients would reject them (IC² Institute, 2025)

Section 5: Emerging and Underreported Use Cases

Agentic AI Workflows

Becker's (2026) documents the emerging shift from ambient AI that records to "agentic AI" that takes action, automatically submitting prior authorization requests, cross-referencing insurance coverage, generating referral documentation, and triggering follow-up workflows. This represents the near-term frontier and is currently in early deployment at large systems.

AI-Assisted Care Coordination

Care transitions between settings (hospital to SNF, SNF to home, ED to behavioral health) are high-risk moments for patient harm and readmission. AI tools that synthesize handoff information, flag missing documentation, and automate transition communications are entering the market but are underrepresented in current prevalence surveys.

Digital Phenotyping for Behavioral Health

Smartphone sensor data (movement patterns, call/text frequency, sleep signals) can generate passive behavioral health monitoring. Currently in research stage but has implications for BH organizations treating high-acuity populations.

SDOH-Aware Predictive Models

A distinct evolution beyond SDOH screening: ML models that integrate social data alongside clinical data to generate risk scores reflecting the full patient picture. Most current risk models are trained on clinical data only; SDOH integration may improve both accuracy and equity when implemented with appropriate data quality validation and governance controls. This is an active research and early-deployment area.

AI for Quality Reporting and Regulatory Compliance

Automating UDS reporting for FQHCs, CMS quality measure documentation, HEDIS measure gap identification, and Joint Commission compliance documentation. This is largely invisible in adoption surveys but represents significant administrative burden that AI tools are beginning to address.

Section 6: What AI Is Not Yet Doing Reliably

For organizations evaluating AI investment, the following represent areas where vendor claims frequently exceed documented performance:

• Autonomous clinical diagnosis: AI-assisted flagging is maturing; autonomous diagnosis without clinician review is not appropriate and not validated in clinical workflows for the org types in scope
• Reliable conversational AI for high-acuity behavioral health: Risk of harm in vulnerable populations is documented; deployment requires governance infrastructure commensurate with clinical risk
• Cross-system model portability: Most AI models perform well within the EHR they were trained on and degrade when applied to different data environments
• Equitable performance across all patient populations: Algorithm bias is documented and unresolved in most commercially deployed tools
• Replacing human judgment in complex social situations: SDOH navigation, care coordination for complex cases, and therapeutic relationships remain human-dependent

Section 7: Report Limitations and Disclosures

Evidence Base Limitations

This report reflects the best available published evidence as of June 2026. Several important caveats apply:

Consulting firm coverage: Reports from EY, Accenture, Deloitte, Leidos, McKinsey, and KPMG that specifically address AI adoption in small and mid-sized healthcare organizations were not identified in the research. Published consulting reports tend to focus on large health systems and generally do not disaggregate findings to the organization-size and type level this report addresses. This is a notable gap in the existing literature, and, for healthcare organizations in these segments, an indication that the landscape analysis specific to their context does not yet exist from traditional advisory sources.

FQHC-specific quantitative data: HRSA does not publish systematic AI adoption surveys for FQHCs. NACHC has not published AI-specific research. FQHC data in this report derives from foundation research and documented case studies rather than national survey data. Readers should treat FQHC prevalence rankings as directional rather than statistically representative.

Behavioral health outcomes data: Most behavioral health AI tools are in early-stage deployment; peer-reviewed outcomes data beyond pilot studies is limited. Vendor claims in this segment should be evaluated with particular care.

Vendor pricing transparency: AI tool pricing is largely opaque in the published literature. Per-seat, per-encounter, and percentage-of-collections pricing models vary significantly and are not publicly documented in ways that allow meaningful comparison. Any procurement process should include direct vendor pricing conversations and a total cost of ownership analysis.

Adoption rate variability: Specific adoption percentages across surveys reflect different methodologies, sample sizes, and definitions of "AI use." Numbers should be treated as directional indicators rather than precise benchmarks.

AI Research Disclosure

This report was developed with the assistance of AI research and synthesis tools. The research process involved:

1. Multi-source search and retrieval: AI-assisted tools conducted parallel searches across five research angles, identifying and prioritizing sources by relevance and methodological quality.
2. Claim extraction: Structured claim extraction was applied to each source, identifying specific assertions and their supporting evidence.
3. Cross-validation: Extracted claims were reviewed using structured validation rules by separate AI processes operating independently of the initial research, designed to identify unsupported assertions, implausible statistics, or internally inconsistent claims.
4. Human editorial synthesis: The final report structure, analysis, interpretations, and recommendations reflect human editorial judgment applied to the validated research output.

Healthcare executives evaluating AI tools, which is precisely what this report covers, should understand that this process models a responsible AI-assisted research workflow: AI for scale and retrieval, structured validation for quality control, and human judgment for interpretation and accountability. We apply the same governance discipline to our own use of AI that we recommend for our clients.

All cited sources are identified by title, publisher, and URL. Readers are encouraged to consult primary sources directly for any finding used in organizational decision-making.

Cited Sources

1. ONC/ASTP Health IT Data Brief No. 80, "Hospital Trends in the Use, Evaluation, and Governance of Predictive AI, 2023–2024." September 2025. https://healthit.gov/data/data-briefs/hospital-trends-use-evaluation-and-governance-predictive-ai-2023-2024/

1. MGMA STAT, "Pace of AI Adoption in Medical Groups Quickens in 2024." Medical Group Management Association, 2024. https://www.mgma.com/mgma-stat/pace-of-ai-adoption-in-medical-groups-quickens-in-2024

1. MGMA/Humana, "AI Adoption in the Value-Based Era." Medical Group Management Association, 2024–2025. https://www.mgma.com/deep-dives/ai-adoption-in-the-value-based-era

1. American Medical Association, "AMA: Physician Enthusiasm Grows for Health Care AI." AMA Press Release, 2024. https://www.ama-assn.org/press-center/ama-press-releases/ama-physician-enthusiasm-grows-health-care-ai

1. HIMSS/Medscape, "AI Adoption by Health Systems." Joint report, 2024. https://www.himss.org/news-center/himss-and-medscape-unveil-groundbreaking-report-ai-adoption-health-systems

1. AHA Center for Health Innovation, "4 Actions to Close Hospitals' Predictive AI Gap." November 2025. https://www.aha.org/aha-center-health-innovation-market-scan/2025-11-04-4-actions-close-hospitals-predictive-ai-gap

1. California Health Care Foundation, "AI Tools Promise Better Care but Challenge Safety-Net Providers." 2025. https://www.chcf.org/resource/ai-tools-promise-better-care-challenge-safety-net-providers/

1. IC² Institute, University of Texas at Austin, "AI in Health Care: Texas Safety-Net Providers Share Perceived Benefits and Barriers." June 2025. https://ic2.utexas.edu/news/new-statewide-study-of-safety-net-providers-reveals-differing-perceptions-about-ai-in-health-care/

1. PMC/NIH, "Adoption of Artificial Intelligence in the Health Care Sector." Peer-reviewed study, 2025. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12639477/

1. JMIR Human Factors, "Barriers to and Facilitators of Artificial Intelligence Adoption in Health Care: Scoping Review." 2024. https://humanfactors.jmir.org/2024/1/e48633

1. JMIR Aging / PMC, "Adoption of AI-Enabled Robots in Long-Term Care Homes by Health Care Providers: Scoping Review." 2024. https://pmc.ncbi.nlm.nih.gov/articles/PMC11387915/

1. npj Digital Medicine (Nature), "Advancing Healthcare AI Governance Through a Comprehensive Maturity Model." February 2026. https://www.nature.com/articles/s41746-026-02418-7

1. Digital Medicine Society (DiMe), "3 Key Insights for the 2026 Health AI Horizon." 2026. https://dimesociety.org/newsroom/blog/3-key-insights-for-the-2026-health-ai-horizon/

1. Social Current, "Artificial Intelligence in Mental Health Care: Promise, Risk, and Responsibility." May 2026. https://www.social-current.org/2026/05/artificial-intelligence-in-mental-health-care-promise-risk-and-responsibility/

1. Commonwealth Fund, "Digital Innovations at CHCs: AI Ambient Scribing in Rural Health Care." 2026. https://www.commonwealthfund.org/blog/2026/digital-innovations-community-health-centers-ai-ambient-scribing-rural-health-care

1. TechTarget Health Tech Analytics, "How an FQHC is Using Ambient AI to Reduce Clinician Burden." Case study (Neighborhood Healthcare / Nabla). https://www.techtarget.com/healthtechanalytics/feature/How-an-FQHC-is-using-ambient-AI-to-reduce-clinician-burden

1. Becker's Hospital Review, "From Ambient AI to Agentic Workflows: What's Ahead for Healthcare in 2026." 2026. https://www.beckershospitalreview.com/healthcare-information-technology/from-ambient-ai-to-agentic-workflows-whats-ahead-for-healthcare-in-2026/

1. Becker's Hospital Review, "From Pilot to Priority: The Rise of Ambient AI Scribes in Healthcare." 2025. https://www.beckershospitalreview.com/healthcare-information-technology/ai/from-pilot-to-priority-the-rise-of-ambient-ai-scribes-in-healthcare/

1. McKnight's Long-Term Care News, "Experts Look Ahead: Technologies That Will Help Long-Term Care Providers in 2025 and Beyond." 2025. https://www.mcknights.com/news/experts-look-ahead-technologies-that-will-help-long-term-care-providers-in-2025-and-beyond/

1. medRxiv (preprint), "AI Implementation in Safety Net Healthcare: Understanding Barriers." April 2026. https://www.medrxiv.org/content/10.64898/2026.04.07.26350351v1.full.pdf