Artificial Intelligence at Mayo Clinic

Mayo Clinic is a nonprofit academic medical center headquartered in Rochester, Minnesota, with additional campuses in Arizona and Florida and a regional health system spanning three Upper Midwest states. The organization employed nearly 85,000 people in 2025 and posted $473 million in operating income. Mayo committed $9 billion in capital investment through its “Bold. Forward. Unbound.” expansion. Independent rankings have named it Newsweek’s No. 1 hospital in the world for the eighth consecutive year.

That scale extends to artificial intelligence. Mayo describes more than 200 AI projects underway across stages of maturity, from early feasibility work to full clinical deployment.

In 2025, Mayo integrated 22 AI-enabled Mayo Clinic Platform solutions into clinical practice and closed nearly 200 new AI, biopharma, and diagnostics agreements.

Since launching the Mayo Clinic Platform in 2019, it has assembled a research data infrastructure spanning more than 15 million patient records and billions of radiology images, lab results, and clinical notes.

This article examines two AI use cases that show how Mayo applies that data and clinical expertise inside its own operations.

AI-Enabled ECG Screening for Early Disease Detection — Detecting asymptomatic heart disease before symptoms appear, using data generated by a routine, low-cost cardiac test.

AI-Powered Chart Review with Record Time — Cutting the hours physicians spend manually reviewing fragmented, unsorted patient files from other health systems, freeing up more time for direct patient care.

We begin by examining how Mayo Clinic applies AI-enabled ECG analysis to address early, asymptomatic detection of heart disease.

AI-Enabled ECG Screening

Asymptomatic left ventricular dysfunction is a precursor to heart failure that standard screening often misses, since diagnosing it has traditionally required an echocardiogram. This test needs specialized equipment and a trained sonographer.

Heart failure now affects nearly 6.7 million Americans, a figure projected to reach 8.7 million by 2030, and cost the U.S. health system an estimated $32 billion in direct medical spending in 2020 alone. Because the underlying dysfunction usually goes undiagnosed until symptoms appear, fewer than one in four eligible patients receive guideline-recommended therapy. When a condition is this costly to treat late and this hard to catch early, the highest-yield AI investment is often earlier detection built from data an organization is already collecting — which is what Mayo’s cardiology researchers set out to do.

Mayo screened more than 625,000 paired ECG and echocardiogram records from its own patients to assemble a study population, then trained a neural network on nearly 98,000 of those pairs to recognize electrical patterns in a standard EKG tied to a weakened heart pump. The approach has since expanded well beyond that original use:

It now also flags atrial fibrillation, cardiac amyloidosis, aortic stenosis, hypertrophic cardiomyopathy, and biological age.

It works across both traditional 12-lead ECGs and single-lead readings from smartwatches and a digital stethoscope.

The amyloidosis version was validated across 25,525 patients at four U.S. health systems, achieving 78.9% sensitivity and 91.2% specificity.

Model quality here tracks the depth of Mayo’s own historical data rather than any particular algorithmic novelty.

The screening step adds no new work for the clinician or the patient: no new test is ordered, since the AI reads the ECG already collected during a routine visit. A positive signal prompts a confirmatory echocardiogram or referral that might not otherwise have been ordered. Portable versions extend that screening beyond the clinic—an AI-enabled digital stethoscope flagged twice as many cases of peripartum cardiomyopathy as standard care in a Nigerian obstetric study, illustrating how AI adoption tends to stick when it rides an existing workflow step rather than adding a new one to remember.

Mayo tested the tool prospectively in the EAGLE trial, enrolling 22,641 patients across 348 primary care clinicians and 45 medical centers in Minnesota and Wisconsin over eight months. The results, published in Nature Medicine:

AI-guided screening increased diagnoses of low ejection fraction by 32% overall compared with usual care — about five additional diagnoses per 1,000 patients screened.

The 12-lead algorithm is now FDA-cleared and licensed to Anumana, a company Mayo co-founded with nference, a healthcare data analytics firm. At the same time, Mayo partnered with Eko Health to develop and commercialize a single-lead version for handheld and wearable devices.

Newer applications, such as amyloidosis detection (cleared by the FDA in April 2026), are still early in commercial rollout.

Screenshot of: Illustration for Anumana’s ECG-AI LEF artificial intelligence (AI) model. (Source: Cardiovascular.com)

Taken together — randomized trial evidence, regulatory clearance, and outside licensing — this is one of Mayo’s most evidence-backed AI applications to date, not an experimental pilot.

AI-Powered Chart Review with Record Time

Ahead of a patient visit, Mayo Clinic physicians often face dozens or even hundreds of pages of medical records to review — a burden compounded by the fact that many patients come to Mayo seeking a third or fourth opinion, arriving with unsorted documents from other health systems. According to Dr. Alexander Ryu, an internal medicine physician and vice chair of innovation for Mayo’s Department of Medicine, the hospital receives tens of millions of pages of records each year, and needed a way to surface the important information buried in that volume.

That volume problem isn’t unique to Mayo. According to a 2025 survey published in the Journal of the American Medical Informatics Association, 77% of the health system leaders surveyed cited immature AI tools as one of the biggest barriers to AI adoption — underscoring why a tool solving a well-defined administrative bottleneck, rather than a broad diagnostic claim, was the more tractable starting point.

Record Time, developed with Scale AI using the Scale Generative AI Platform, ingests fragmented external patient records and organizes them chronologically, generating concise summaries and making the material searchable. According to Scale AI, the collaboration’s broader scope also includes automating detection of safety events — such as wrong-site surgeries or falls —hidden within routine reporting noise, and helping staff spend less time on administrative tasks. Data used in the collaboration remains within Mayo Clinic’s HIPAA-compliant environment.

Physicians spend less time manually assembling a patient’s history before a visit and more time on direct interaction, illustrated by two data points:

Physicians spend less time manually assembling a patient’s history before a visit and more time on direct interaction.

According to Scale AI, physicians have gained an average of 11 more minutes with each patient since the tools launched, while maintaining an expert standard of care.

Dr. Ryu separately described time savings of five to thirty minutes of prep per visit, depending on case complexity, and said the tool also helps him avoid missing details buried in a file that could affect treatment or testing decisions.

This is a deployed, in-use tool, not a pilot — one of roughly 150 AI models Mayo now has running across the system, giving it a track record beyond a single department or trial.