We’re putting $5.5M behind research and development of smaller, specialized AI models for real-world deployment — Read our manifesto →
    All posts

    [Case Study] 4,000 Clinicians in 16 Weeks: Cleveland Clinic's Ambient AI Scaling Playbook

    How governance, mandatory training, specialty waves, live support, and utilization metrics turned an AI scribe into enterprise clinical infrastructure.

    Conscious Engines

    Industry: Healthcare
    Organization: Cleveland Clinic
    Use case: Enterprise-scale ambient clinical documentation
    Evidence basis: 2026 npj Health Systems implementation report
    Disclosure: This is an independent analysis by Conscious Engines. The paper was authored through a health-system and vendor partnership, and operational data are not publicly available for independent replication.

    1. Outcome at a Glance

    Cleveland Clinic trained and onboarded more than 4,000 ambulatory clinicians in under 16 weeks, representing approximately 80% of its ambulatory clinicians. More than 12 months after enterprise deployment, over 4,800 clinicians had used the tool across more than 3.5 million encounters.

    Key Outcomes

    More than 4,000 clinicians

    Initial onboarding

    Completed in under 16 weeks reported in the cited case.

    About 80%

    Share of ambulatory clinicians

    Broad organizational reach reported in the cited case.

    More than 4,800

    Users after 12 months

    Continued expansion reported in the cited case.

    More than 3.5 million

    Encounters after 12 months

    Production scale reported in the cited case.

    70%

    Encounter-level use among established users

    Stronger than licence activation reported in the cited case.

    MeasureReported resultWhy it matters
    Initial onboardingMore than 4,000 cliniciansCompleted in under 16 weeks
    Share of ambulatory cliniciansAbout 80%Broad organizational reach
    Users after 12 monthsMore than 4,800Continued expansion
    Encounters after 12 monthsMore than 3.5 millionProduction scale
    Encounter-level use among established users70%Stronger than licence activation
    Net Promoter Score60Clinician advocacy
    Customer satisfaction score96.6%User-reported satisfaction
    Users saying tool increased likelihood of staying in practice60%Retention-related perception
    Rollout support inquiriesMore than 900Support demand
    Average support response2 minutesOperational support performance

    The most important number is 70% encounter-level utilization among established users, defined as clinicians who had used the tool at least 50 times. Many reports call a person “adopted” after a single use. Encounter-level utilization shows whether the tool is present in normal clinical work.

    2. The Operational Problem

    Cleveland Clinic is a large academic health system with 23 hospitals, 276 outpatient locations, and about 80,000 caregivers worldwide. At this scale, a strong model can still fail through training bottlenecks, specialty mismatch, poor support, EHR friction, or unclear ownership.

    Ambient documentation also varies by clinical workflow. Primary care may produce repeatable conversational notes. A complex specialty may require unique templates, examination language, multiple participants, or structured findings. A single rollout date does not address that variation.

    The system needed to convert a selected product into an operating service with:

    • clear executive and day-to-day governance;
    • EHR and mobile-device support;
    • specialty-specific training and configuration;
    • rapid issue resolution;
    • revenue-cycle measurement;
    • continuous feedback and product changes;
    • transparent adoption and utilization reporting.

    3. What Was Built

    Cleveland Clinic and Ambience Healthcare built a joint deployment organization around the ambient system.

    System at a Glance

    Executive Sponsor Group

    Direction, guardrails, accountability, and barrier removal.

    Project Operating Council

    Day-to-day decisions and cross-functional coordination.

    IT group

    EHR build, mobile-device management, maintenance, and troubleshooting.

    Revenue-cycle group

    Financial and coding indicators.

    Training and communications

    Onboarding, education, and departmental engagement.

    Clinical product group

    User feedback and product priorities.

    Operating groupResponsibility
    Executive Sponsor GroupDirection, guardrails, accountability, and barrier removal
    Project Operating CouncilDay-to-day decisions and cross-functional coordination
    IT groupEHR build, mobile-device management, maintenance, and troubleshooting
    Revenue-cycle groupFinancial and coding indicators
    Training and communicationsOnboarding, education, and departmental engagement
    Clinical product groupUser feedback and product priorities
    Outreach and monitoringUtilization and adoption support

    The rollout used waves. Prior pilot users entered first. Approximately 80% of target users, especially primary care, entered the next wave. Additional specialties followed, with complex specialties placed later while model maturity and workflow were improved.

    The organization also created direct, 24-hour support with reciprocal handoffs between vendor and internal IT teams. Inquiries included physical-exam templates, multi-clinician settings, content corrections, and system policies. These are production-model issues, not generic help-desk questions.

    4. How It Reached Production

    The implementation report identifies four practices that supported scale.

    Contract around adoption milestones. Vendor and health-system incentives were connected to the number of clinicians onboarded in each wave.

    Require training before access. Live virtual sessions ran three times daily, supported by physician leaders. Department meetings and self-directed modules filled scheduling gaps.

    Sequence by complexity. Complex specialties followed after more mature workflows, rather than receiving an immature universal template.

    Support clinicians in real time. More than 900 inquiries received an average two-minute response during rollout. Rapid support turned friction into product feedback.

    Give leaders utilization dashboards. Department chairs could see onboarding and use, identify gaps, and adjust outreach.

    A bespoke deployment can apply the same operating model while giving the health system more control over specialty templates, medical ASR, model routing, private hosting, and evaluation data. The product must be adapted to the clinical workflow, not only taught through training.

    5. What Healthcare Leaders Should Take Away

    Cleveland Clinic's case shows that enterprise AI scale is an organizational capability. The model generated notes, but governance, training, support, feedback, and specialty sequencing produced durable use.

    A healthcare rollout should track four different metrics: licence activation, clinician adoption, encounter-level utilization, and accepted-note rate. Only the last two show whether the system repeatedly creates usable output.

    The bespoke opportunity is to own the layers that determine those measures: specialty speech models, templates, RAG, EHR integration, evaluation, support telemetry, and model routing. The objective is not to maximize model calls. It is to produce a safe note that clinicians choose to use.

    Sources