Hippocratic AI

Hippocratic AI is a Palo Alto-based company that builds patient-facing voice agents for healthcare organizations. Founded in 2023 by Munjal Shah, the company's core premise is that a global shortage of healthcare workers can be partially addressed by AI agents that handle high-volume, low-acuity patient communication tasks: post-discharge follow-up calls, chronic care check-ins, medication adherence reminders, preventive care outreach, and appointment scheduling. The agents are explicitly non-diagnostic and non-prescriptive, designed to operate within the scope of what a trained clinical support staffer would do, not to replace physicians or make clinical judgments. By May 2026, Hippocratic AI has raised $404 million across three funding rounds, reached a $3.5 billion valuation at its November 2025 Series C led by Avenir Growth, and counted more than 1,000 healthcare organizations among its partners, including WellSpan Health, OhioHealth, University Hospitals, Cincinnati Children's, and Universal Health Services.

The platform runs on Polaris, a proprietary constellation architecture. As of April 30, 2026, the current version is Polaris 5.0: a 5-trillion-parameter system anchored by a 700-billion-parameter core model and surrounded by specialized support models for drug safety, clinical escalation, lab interpretation, insurance benefits lookup, and cough/acoustic detection for respiratory assessment. The system achieves 1.5-second time-to-first-audio latency and supports mid-call language switching between English, Spanish, and Mandarin. Clinical safety benchmarks published at the Polaris 5.0 launch show 99.95% drug safety accuracy and 99.75% clinical escalation accuracy, outperforming GPT-5 Mini and Gemini 2.5 Flash on the same tasks. Two main products are available: AI Front Door for end-to-end patient communication workflows, and Nurse Co-Pilot for AI-assisted nursing team operations. A Healthcare AI Agent App Store, launched in January 2025, allows health systems to co-develop custom agents for specific patient populations and conditions through Hippocratic's API.

What Hippocratic AI actually is in May 2026

Hippocratic AI occupies a specific, bounded lane in healthcare AI: the patient communication layer. When a patient is discharged after a joint replacement, the AI agent calls within 48 hours to run a structured symptom check, assess medication adherence, and escalate to a human nurse if something sounds clinically concerning. When a Medicare Advantage plan needs to reach 50,000 members overdue for colorectal cancer screening, agents make outbound calls at scale, answer common questions, and schedule appointments. When a pharma company is running a clinical trial, agents handle patient check-ins and protocol adherence across the trial population.

The Polaris constellation architecture is what distinguishes Hippocratic from deploying a general-purpose model fine-tuned on healthcare data. Instead of a single model handling all tasks, Polaris 5.0 uses a primary conversational agent supervised in real time by a ring of specialized evaluator models, each checking for specific failure modes: hallucinated drug information, missed escalation triggers, HIPAA authentication failures, empathy drift. The system has processed more than 180 million patient interactions since commercialization with a reported 99.90% correct clinical guidance rate and zero documented severe harm events.

"The safety information on Anna (the name WellSpan Health gave its Hippocratic AI agent) as our AI care assistant is absolutely terrific. And every one of our clinical leaders and chief medical officers who've reviewed that has been very impressed." -- Roxanna Gapstur, PhD, RN, President and CEO of WellSpan Health, November 2025

The January 2025 Series B ($141 million, led by Kleiner Perkins) brought Hippocratic to unicorn status at a $1.64 billion valuation and launched the Healthcare AI Agent App Store alongside it. The November 2025 Series C ($126 million, led by Avenir Growth) more than doubled the valuation to $3.5 billion, with CapitalG (Google's growth fund), General Catalyst, Andreessen Horowitz, Kleiner Perkins, Cincinnati Children's, and WellSpan Health all participating. Total funding is $404 million. The company has since expanded internationally, with partnerships at Sheba Medical Center in Israel and EUCALIA Inc. in Japan (deploying the first non-diagnostic, patient-facing Japanese generative AI healthcare agent, announced May 2025).

Where Hippocratic AI sits versus Suki AI and Glass Health

The comparison that comes up most often is with Suki AI, but the two tools operate on different sides of the patient-clinician relationship and rarely compete for the same budget. Suki is an ambient scribing platform for physicians and clinical staff. It listens to clinician-patient encounters and generates structured SOAP notes, H&P documents, and progress notes, then pushes them into Epic, Cerner, athenahealth, and MEDITECH via deep bidirectional integrations. Suki's user is the clinician finishing a note after an appointment. Hippocratic AI's user is the patient receiving a follow-up call two days after discharge. A health system can run both without overlap, which is why Suki's investors (Google, among others) and Hippocratic's investors (Kleiner, a16z) are not in zero-sum competition.

The more revealing comparison is with Glass Health, a smaller company building AI clinical decision support. Glass Health listens to clinician-patient encounters and constructs a real-time differential diagnosis (Most Likely, Expanded, Can't Miss categories), assessment and plan generation, and clinical Q&A -- tools for the physician to think better during the encounter. Glass Health integrates with Epic, eClinicalWorks, and Athena via SMART on FHIR and offers a freemium model starting at $29/month, making it accessible to independent clinicians and small practices. Hippocratic AI does not touch differential diagnosis at all and is enterprise-only with custom pricing. Glass Health is also much earlier stage, still in beta for several features, and has not published the parameter count or architecture details of its underlying model. The fundamental difference is focus: Glass Health improves clinical reasoning; Hippocratic AI scales patient communication.

For context on adjacent tools in the healthcare AI stack, Abridge competes directly with Suki in ambient scribing for clinicians, while Nabla and OpenEvidence address clinical knowledge retrieval and diagnostic support, respectively -- all operating in the clinician-facing layer that Hippocratic AI explicitly does not occupy.

How the AI actually works inside Hippocratic AI

The Polaris constellation model is not a fine-tuned version of GPT-4 or Claude or any publicly available model. It is a proprietary multi-model system. Polaris 5.0 contains more than 20 specialized clinical sub-models running alongside the conversational core: a drug safety model checking every medication mention for accuracy, a clinical escalation model monitoring for seven body systems and specific crisis indicators including suicidal ideation and child protective service triggers, a lab interpretation model, an insurance benefits lookup model, and a custom clinical ASR (automatic speech recognition) system optimized for medication pronunciation. The system also includes cough and acoustic detection for respiratory assessment, a capability added in the Polaris 5.0 release.

Before any agent goes live with patients, Hippocratic runs it through a testing protocol involving thousands of licensed U.S. nurses and doctors role-playing as patients. More than 7,500 licensed clinicians have participated in this validation process. Patient satisfaction scores tracked across model versions show consistent improvement: from 8.72/10 with Polaris 2.0 to 8.95/10 with Polaris 3.0 (released March 2025, 4.2 trillion parameters across 22 specialized LLMs). The company has received HITRUST e1 certification (July 2025) for data security compliance.

"At OhioHealth, safety is the cornerstone of every care decision we make, and we hold our partners to the same standard. Hippocratic AI's unwavering commitment to clinical safety, combined with their rigorous testing and validation process, gives us the confidence to deploy their agents in ways that truly support patients and clinicians alike." -- Dr. Michael Ezzie, MD, SVP and President of OhioHealth Physician Group, November 2025

The friction users keep raising

The most substantive ongoing criticism of Hippocratic AI does not come from patients -- it comes from healthcare IT analysts and implementation teams who find that the platform solves a narrower problem than health systems initially expect.

A recurring frustration in analyst coverage: what the AI agents do best (patient education calls, preventive care outreach, post-discharge follow-up) is not actually where nurses spend most of their time. A McKinsey study on nursing workflows found that nurses' biggest time demands are medication administration, locating patients and equipment, and updating documentation -- none of which Hippocratic AI agents can address. The mismatch between "AI reduces nursing burden" in marketing materials and the operational reality of what nurses spend time doing has created some disappointment at the deployment stage.

A second friction point is benchmarking transparency. The Polaris model is fully proprietary. All accuracy claims -- 99.95% drug safety accuracy, zero severe harm events -- are based on internal testing. As of mid-2026, no peer-reviewed journal has published an independent clinical validation of the Polaris system's accuracy in real-world deployment. The company's arXiv papers describe the architecture and methodology, but reproduction is impossible without access to the model.

Third: CMS billing codes. Chronic care management reimbursement under codes 99490 and 99497 requires documented human clinician time. AI agent interactions do not qualify, limiting the revenue case some health systems built into their business justification for the platform.

Fourth: EHR integration depth varies significantly by health system. Agents can pull patient data from EHRs to personalize calls and can document summaries back, but the depth of two-way integration requires substantial technical work on the health system side to configure, and the burden falls primarily on the deploying organization's IT team.

Who Hippocratic AI is for

Hippocratic AI is a strong fit for large health systems, Medicare Advantage plans, and pharma companies running patient engagement programs at scale. The platform makes the most economic sense when an organization has: (a) high volumes of patients needing routine outreach that currently falls through the cracks due to nurse capacity constraints, (b) a meaningful existing technology infrastructure (EHR, phone system, data feeds) to integrate with, and (c) an enterprise IT team able to configure and maintain the integration. Reported outcomes from live deployments include a 30% reduction in 30-day readmission rates for post-discharge follow-up programs, a 360% increase in team capacity for chronic care management at Cincinnati Children's, and 60% of outbound flu shot outreach calls resulting in documented vaccination goals.

Skip Hippocratic AI if you are a small or independent practice without enterprise IT resources, a behavioral health or pediatric specialty practice (the platform explicitly excludes mental health crisis management and patients under age two from its scope), or an organization primarily looking for clinical documentation software. For ambient scribing and clinician-facing AI, Suki AI or Abridge are the relevant choices. For clinical decision support during patient encounters, Glass Health addresses that workflow. Hippocratic AI's value is specifically in the patient outreach layer, and organizations that need it most are those managing large patient populations with limited nursing bandwidth for routine follow-up and preventive care outreach.