iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
AI is already being used in clinical workflows for drafting visit notes, flagging possible sepsis, and supporting heart-failure detection. The documented examples below include 12 named health-system or project settings. They are a curated set, not a complete census—and the available evidence does not establish a verified list of exactly 18 comparable deployments.
That distinction matters: a tool can be announced as available, actively used by clinicians, or evaluated in a care pathway. Those statuses—and the results measured—are not interchangeable.
What counts as a real clinical AI deployment?
For this article, a deployment means that a named health system or project reports introducing or using an AI-enabled tool in a clinical workflow. The examples include a mix of availability, measured use, and evaluation; the table labels the status each source supports. A rollout announcement alone does not establish how many clinicians used a tool or whether it improved patient outcomes.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
NHS England’s account of its AI in Health and Care Award distinguishes first prospective deployment, multisite deployment, and real-world evaluation. Those stages help explain why an evaluation project should not automatically be described as routine clinical use. NHS England’s implementation and evaluation account also identifies safety, accuracy, effectiveness, value, fit with sites, implementation, feasibility of scaling, and sustainability as evaluation domains.
Where are health systems using AI now?
Ambient documentation is the most prominent cluster in these examples. Other reported applications include sepsis prediction and primary-care support for detecting possible heart failure.
| Health system or project | Clinical task and setting | What the source reports | How to read the evidence |
|---|---|---|---|
| Kaiser Permanente / Abridge | Ambient documentation in hospitals and medical offices | In August 2024, Kaiser Permanente announced availability at 40 hospitals and more than 600 medical offices. Its description says patients consent and clinicians review the generated notes. Kaiser Permanente’s announcement | These are locations where the tool was available, not counts of active users or measured patient outcomes. The system’s description of the rollout as the largest of its kind is Kaiser’s characterization, not an independent ranking. |
| Vanderbilt University Medical Center | Ambient scribing for ambulatory and emergency-department clinicians | Enterprise access began January 15, 2025, for more than 2,400 clinicians. By March 31, 2025, 1,223 had used the system; it was used in 20.1% of visit notes during the study’s final week. Vanderbilt’s deployment report | Access, clinician uptake, and share of notes are different measures. The 20.1% figure refers to visit notes in one study week, not all eligible visits or clinicians. |
| Cleveland Clinic / Ambience Healthcare | Ambient documentation for ambulatory clinicians | Implementation began March 10, 2025, and reached more than 4,000 ambulatory clinicians within four months, according to a 2026 implementation report. Cleveland Clinic’s rollout report | The reported scale documents implementation, not clinical benefit by itself. |
| National University Health System (NUHS), Singapore / MediVoice | In-house multilingual ambient scribing | NUHS deployed MediVoice in September 2024. The account says local data requirements influenced its decision to build the tool in-house. NUHS deployment coverage | This is a local implementation choice; it should not be generalized as a requirement or policy for other health systems or jurisdictions. |
| Cleveland Clinic hospitals / TREWS | Sepsis prediction and clinical alerts | The Stanford AI Index 2026 describes the Johns Hopkins-developed, Bayesian Health-commercialized system as deployed across 13 Cleveland Clinic hospitals. It summarizes cited results including an 18.7% relative reduction in sepsis mortality. Stanford AI Index 2026, medicine chapter | The percentage is a relative reduction reported in the Index’s summary of a specific deployment result, not a guaranteed effect for other hospitals. It is not an 18.7-percentage-point absolute reduction. |
| UC San Diego Health / COMPOSER | Deep-learning sepsis monitoring | The Stanford AI Index says the model monitored more than 150 variables per patient and summarizes results across 6,217 admissions: a reported 17% relative mortality reduction, equivalent to 1.9% absolute, with estimated lives saved. Stanford AI Index 2026, medicine chapter | These are results summarized for the reported study and admission population, not a general effect estimate for sepsis AI. |
| Sharp HealthCare | Ambient clinical documentation | The Stanford AI Index reports an 83% reduction in note-writing effort and a 3.5%–6% increase in work relative value units per encounter. Stanford AI Index 2026, medicine chapter | These are source-specific reported measures; note-writing effort and work RVUs are not patient health outcomes. |
| University of Chicago Medicine | Ambient documentation and clinician workflow | The Stanford AI Index reports a 47% reduction in cognitive load and a 58% increase in undivided patient attention. Stanford AI Index 2026, medicine chapter | These reported measures concern clinician experience and attention, not clinical outcomes such as mortality or complications. |
| MaineHealth | Ambient clinical documentation | The Stanford AI Index reports 23% less time spent on clinical notes and use in 70.3% of encounters. Stanford AI Index 2026, medicine chapter | The share refers to the source’s definition of use and its observation period; it is not a universal adoption rate. |
| Northwestern Medicine | Ambient documentation for physicians | For physicians who used the tool in more than half of encounters, the Stanford AI Index reports 11.3 additional patients per month, 24% less documentation time, and a reported 112% return on investment. Stanford AI Index 2026, medicine chapter | The results apply to a subgroup of physicians, and the return figure is a reported calculation—not a guaranteed financial result for another institution. |
| Stanford Health Care | Ambient documentation in outpatient clinics | The Stanford AI Index summarizes a prospective study of 48 physicians that found median time savings of 20 minutes per half-day clinic, along with statistically significant reductions in task load and burnout. Stanford AI Index 2026, medicine chapter | The sample and prospective study context are important; the median time saving is not an estimate for every clinician or clinic. |
| NHS TRICORDER / Eko-DUO | AI-enabled stethoscope evaluated for heart-failure detection support in primary care | NHS Digital reports more than 200 participating GP practices by February 2024. The project timeline runs from January 2023 to August 2025. NHS Digital’s TRICORDER case study | This was a project evaluating the technology. The case study’s £2,400 per-patient and £100 million nationwide savings figures are projections, not reported realized savings. |
Do these deployments show that AI improves patient outcomes?
Some reports describe measured clinical outcomes, while many describe availability, adoption, documentation time, clinician workload, or financial calculations. Those are different kinds of evidence. For example, the TREWS and COMPOSER figures above concern sepsis mortality in specific reported settings; the ambient-scribe examples mainly describe documentation or workflow measures.
A reported association or before-and-after result should not be treated as proof that the AI tool alone caused the change. The study design, patient population, comparison group, follow-up period, and local workflow all affect what a result can establish. The Stanford AI Index is a secondary synthesis of underlying studies, while health-system announcements and implementation reports describe their own deployments; each is useful, but neither makes unlike systems directly comparable.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →For a consequential decision, look for the underlying study and ask whether it reports absolute as well as relative effects, how many patients or encounters were included, and whether the result was independently evaluated. A rollout count cannot answer those questions.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Which AI projects are evaluations rather than established routine deployments?
The NHS AI in Health and Care Award illustrates the difference between funding, evaluation, and routine use. NHS England reports that the 2020–2024 programme allocated more than £100 million to support design, development, and deployment. It also says 13 Phase 4 technologies were independently evaluated for multisite deployment and real-world outcomes. Those figures describe programme support and evaluation activity—not 13 proven products in routine care or proof of effectiveness. NHS England’s programme account discusses evaluation contexts including stroke imaging and chest-image AI, as well as practical implementation challenges.
Generative AI workflow reports need the same care. The Stanford AI Index summarizes ChatEHR usage of 23,000 sessions among 1,075 trained users over three months, along with tools that help explain patient records or test results. Those figures describe workflow use in the reported context; they do not establish autonomous diagnosis or routine deployment across health systems. Stanford AI Index 2026, medicine chapter
Quick Recap
What should a hospital check before adopting a clinical AI tool?
- Define the task. Specify whether the system drafts documentation, flags risk, interprets images, or supports a diagnostic pathway; performance in one task does not establish suitability for another.
- Separate access from actual use. Track eligible staff, active users, encounters where the system was used, and sustained use over time as separate measures.
- Keep the clinician’s role explicit. Establish who reviews, edits, accepts, or acts on generated output, and how the workflow handles errors or uncertainty.
- Evaluate the right outcomes. Measure safety and accuracy alongside effectiveness, value, fit with local sites, implementation burden, scalability, and sustainability—the domains NHS England identifies for real-world evaluation.
- Report denominators and time periods. A percent change without its population, baseline, comparison, and observation window can be misleading.
- Plan for ongoing monitoring. A result from one site or period does not ensure performance will persist after workflows, staffing, or patient populations change.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
