By Sagar Shankaran, Founder of CallSphere
Mayo Clinic's year-long pilot shows AI agents handling patient intake, preliminary diagnosis, and insurance pre-authorization dramatically cut wait times while maintaining diagnostic accuracy.
Key takeaways
Mayo Clinic has published the results of a year-long pilot program deploying AI agents across patient intake, preliminary diagnostic assessment, and insurance pre-authorization workflows. The results, published in the New England Journal of Medicine on March 13, 2026, show a 60% reduction in average diagnostic wait times, from 23 days to 9.2 days, for patients in the pilot cohort. The study has been hailed as the most rigorous evidence to date that AI agents can meaningfully improve healthcare delivery at scale.
The pilot, conducted across Mayo Clinic's Rochester, Minnesota, and Phoenix, Arizona campuses from March 2025 to February 2026, enrolled 47,000 patients across cardiology, oncology, neurology, and general internal medicine departments. The AI agent system, developed in partnership with Epic Systems and Google Health, operated alongside existing clinical workflows rather than replacing them.
Mayo's system deploys three specialized AI agents that work in sequence to accelerate the patient journey from initial contact to diagnosis:
flowchart LR
CALLER(["Patient or Caregiver"])
subgraph TEL["Telephony"]
SIP["Twilio SIP and PSTN"]
end
subgraph BRAIN["Healthcare AI Agent"]
STT["Streaming STT<br/>Deepgram or Whisper"]
NLU{"Intent and<br/>Entity Extraction"}
TOOLS["Tool Calls"]
TTS["Streaming TTS<br/>ElevenLabs or Rime"]
end
subgraph DATA["Live Data Plane"]
CRM[("CRM and Notes")]
CAL[("Calendar and<br/>Schedule")]
KB[("Knowledge Base<br/>and Policies")]
end
subgraph OUT["Outcomes"]
O1(["Appointment booked"])
O2(["Prescription refill request"])
O3(["Triage to clinician"])
end
CALLER --> SIP --> STT --> NLU
NLU -->|Lookup| TOOLS
TOOLS <--> CRM
TOOLS <--> CAL
TOOLS <--> KB
NLU --> TTS --> SIP --> CALLER
NLU -->|Resolved| O1
NLU -->|Schedule| O2
NLU -->|Escalate| O3
style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
style O1 fill:#059669,stroke:#047857,color:#fff
style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937
When a patient calls to schedule an appointment or is referred by an external provider, the intake agent conducts an initial conversation, by phone (using voice AI) or through Mayo's patient portal. The agent collects symptoms, medical history, current medications, and relevant lifestyle factors through a structured but conversational interview.
Critically, the intake agent does not simply fill out a form. It uses clinical reasoning to ask follow-up questions based on symptom patterns. If a patient mentions chest pain, the agent probes for duration, character, radiation, exacerbating factors, and associated symptoms, following clinical assessment protocols derived from Mayo's own diagnostic guidelines.
The intake agent then generates a structured clinical summary and a preliminary problem list, which is routed to the appropriate department. In the pilot, this process reduced intake processing time from an average of 4.2 days (the time between initial contact and a completed intake form reaching a physician) to 0.3 days.
Once the intake summary is complete, the diagnostic triage agent analyzes the patient's information against Mayo's clinical knowledge base, relevant medical literature, and anonymized patterns from Mayo's database of over 10 million patient records. The agent generates a ranked differential diagnosis, a list of possible conditions ordered by likelihood, along with recommended diagnostic tests.
This is where the system's value becomes most apparent. Previously, a patient might wait days or weeks for an initial physician appointment, only to have the physician order tests that require additional wait times. The triage agent pre-identifies likely needed tests, allowing the scheduling system to book the patient's first physician appointment and diagnostic tests in a coordinated sequence.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent for healthcare in your browser — 60 seconds, no signup.
For a patient with suspected cardiac issues, for example, the triage agent might recommend an ECG, echocardiogram, and blood panel, and the scheduling system would book all three on the same day as the physician consultation. Previously, these would often be scheduled sequentially over multiple visits.
The triage agent's diagnostic suggestions proved remarkably accurate. In a blinded comparison, the agent's top-three differential diagnosis matched the eventual confirmed diagnosis 84.7% of the time, compared to 82.1% for the initial differential generated by physicians alone. The agent's performance was strongest in cardiology (89.2% concordance) and weakest in neurology (78.4%), a finding the researchers attributed to neurology's greater reliance on physical examination findings that the agent cannot assess.
In the U.S. healthcare system, insurance pre-authorization is one of the most significant bottlenecks in patient care. Obtaining approval for diagnostic tests and procedures requires submitting detailed clinical justification to insurance companies, a process that typically takes 3-14 business days and consumes substantial physician and staff time.
Mayo's authorization agent automates this process by generating pre-authorization requests that include the specific clinical documentation, diagnostic codes, and supporting evidence that insurance companies require. The agent submits requests electronically and monitors their status, escalating to human staff only when a request is denied or requires additional clinical input.
In the pilot, the authorization agent reduced average pre-authorization time from 7.3 days to 1.8 days. Approval rates actually increased slightly, from 91% to 94%, because the agent's submissions were more consistently complete and properly coded than manual submissions.
The study's safety data is perhaps its most important finding. Across 47,000 patients, the researchers identified zero cases where the AI agent system caused a clinically significant diagnostic delay or error that would not have occurred under standard workflows. Two cases were flagged where the triage agent's differential diagnosis omitted a condition that was later diagnosed, but in both cases, the physician's evaluation caught the omission before any patient harm occurred.
Patient satisfaction scores in the pilot cohort were significantly higher than the control group. 87% of pilot patients rated their experience as "excellent" or "very good," compared to 71% in the control group. The most commonly cited reason was reduced waiting time and the feeling that "the system already understood my situation" when they arrived for their first appointment.
Physician satisfaction was also positive, though more nuanced. 73% of physicians in the pilot reported that the AI intake summaries and diagnostic suggestions were "very useful" or "somewhat useful." However, 22% expressed concern about "automation complacency," the risk that physicians might rely too heavily on the agent's diagnostic suggestions rather than conducting their own independent assessment.
Dr. John Halamka, president of the Mayo Clinic Platform, addressed this concern directly: "We designed the system to inform, not replace, clinical judgment. The AI agent's output is presented as one input among many, alongside the patient's own account, physical examination findings, and the physician's expertise. We monitor regularly for signs that physicians are deferring to the agent rather than thinking independently."
Still reading? Stop comparing — try CallSphere live.
See the healthcare AI agent handle a real call — complete, industry-specific, and live in your browser. No signup.
The economic implications of the pilot are substantial. Mayo estimates that the AI agent system reduced per-patient administrative costs by $340, primarily through reduced staff time on intake processing, scheduling coordination, and insurance authorization. Across the pilot cohort of 47,000 patients, this represents approximately $16 million in annual savings.
More significantly, the 60% reduction in diagnostic wait times allowed Mayo to increase patient throughput without adding clinical staff. The Rochester campus saw a 12% increase in new patient appointments during the pilot period, generating additional revenue while improving the patient experience.
These figures have attracted intense interest from the broader healthcare industry. HCA Healthcare, the nation's largest hospital chain, announced on March 15 that it would begin a similar pilot across five of its facilities. Kaiser Permanente, Cleveland Clinic, and Johns Hopkins have all indicated interest.
The Mayo pilot operated under an FDA regulatory framework established in 2025 for AI clinical decision support systems. The triage agent is classified as a Class II medical device requiring 510(k) clearance, which Mayo and Google Health obtained in early 2025.
The study raised several ethical questions that the researchers addressed directly:
Health equity: The researchers found no statistically significant differences in agent performance across racial, ethnic, or socioeconomic groups in the pilot cohort. However, they acknowledged that the training data is heavily weighted toward Mayo's patient population, which is not representative of the broader U.S. population, and cautioned against assuming equity in other settings.
Data privacy: Patient data used by the agents is processed within Mayo's own infrastructure and is not shared with Google or any external party. The models were fine-tuned on Mayo's data within a secure environment, and the resulting weights remain Mayo's property.
Informed consent: All pilot participants provided informed consent, including the right to opt out of AI agent processing at any time. Approximately 3% of eligible patients declined to participate.
The Mayo Clinic study provides the strongest evidence yet that AI agents can be safely and effectively deployed in clinical settings. The 60% reduction in diagnostic wait times is not just a metric of operational efficiency; for patients waiting for cancer diagnoses or cardiac evaluations, faster diagnosis can mean earlier treatment, better outcomes, and reduced anxiety.
However, scaling this approach beyond well-resourced academic medical centers will be challenging. Mayo's success depends on its extensive digital infrastructure, deep clinical knowledge base, and willingness to invest in rigorous evaluation. Smaller hospitals and community health centers may lack these resources.
The healthcare industry is watching closely. If the Mayo model can be replicated at other institutions, AI agents could fundamentally reshape how healthcare is delivered in the United States and beyond, reducing the administrative burden that consumes an estimated 34% of U.S. healthcare spending, while improving the speed and quality of patient care.
Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
AWS HealthScribe became the open scribe layer EHR vendors built on top of in 2026. Here's the API surface, the per-encounter pricing, the BAA terms.
Apollo, Manipal, and Narayana scaled AI agents across Bangalore in 2026. Here's the deployments across radiology, intake, and follow-up, the costs.
Notable's AI agents now handle scheduling, intake, and revenue cycle for 6,000+ clinics in 2026. Here's the multi-agent architecture, the per-clinic pricing.
Enterprise CIO Guide perspective on Hippocratic AI's deployment numbers show healthcare voice agents are moving from pilot to production across major US health systems.
Abridge raised $250M in April 2026 at a $2.7B valuation. We break down the deployment numbers, the EHR integrations across Epic and Cerner. The Q2 2026 buyer briefing.
Home health agencies use AI voice agents for daily check-ins on 2.5M+ patients in 2026. Here's the deployments at major agencies, the per-patient pricing.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco