Safety-first conversational AI agents and 30+ agentic orchestrators for payers, providers, and life sciences — built on the patented Polaris constellation, 250M+ patient interactions, $3.5B Series C
HQ: Menlo Park, CA (relocated from Palo Alto) | Founded: 2023 | Employees: ~235 (Crunchbase band 101-250) | Stage: Series C — $126M at $3.5B valuation (Nov 2025); ~$444M total raised per company site (Aug 2026) | Website: https://www.hippocraticai.com
Hippocratic AI was founded in May 2023 by serial AI entrepreneur Munjal Shah (Like.com — sold to Google; Andale — sold to Alibaba; Health IQ) together with generative-AI researchers, hospital executives, physicians, and Medicare experts, with the explicit thesis that generative AI is not yet safe enough inside a physician's scope of practice — so their agents would handle non-diagnostic, patient-facing clinical work first. From day one the design constraint was safety parity with human clinicians, enforced through the Polaris Safety Constellation: a primary language model (~700B parameters in the 2026 generation) whose every call is monitored by 30+ specialist supervising models for escalation needs, medication recognition, adverse-event detection, and drug-safety checks. The company raised a $50M seed from General Catalyst and Andreessen Horowitz at launch (May 2023), published the Polaris architecture (arXiv, March 2024), launched its first commercial generative-AI healthcare agent in June 2024, and signed 23 health systems, payors, and pharma clients in 23 weeks. It received its first US patent and claimed clinician-level safety parity in January 2025 alongside an agent app store where licensed clinicians design and earn from agents. After a $141M Kleiner-Perkins-led Series B at $1.64B (Jan 2025) and a $126M Avenir-Growth-led Series C at $3.5B (Nov 2025), the company crossed 50+ clients in 6 countries and 1,000+ use cases. In Aug 2026 it shipped the next phase — Agentic Orchestrators — 30+ teams of coordinated voice agents aimed at outcomes (readmission reduction, Star Ratings/HEDIS improvement, chronic care, trial enrollment, pharmacy, rapid-response climate outreach) rather than single tasks, and the flagship WellSpan-deployed agent 'Ana' now handles 160,000+ calls/month. Hippocratic AI's agents never prescribe or diagnose, a boundary the company treats as immutable.
Hippocratic AI is the category leader in patient-facing clinical AI — the hardest, most liability-saturated corner of agentic AI — and it got there by applying a healthcare-native safety doctrine rather than by racing raw model size. The durable moat is the Polaris constellation architecture (a ~700B-param primary model governed by 30+ supervising specialist models) plus a closed human-validation flywheel: 250M+ real patient interactions, 7,700+ US-licensed clinicians shaping and certifying agents, and health systems (WellSpan, UHS, Cincinnati Children's, HonorHealth, Memorial Hermann) as both customers and investors. The company has deliberately walled itself off from diagnosis and prescribing — a regulatory and liability boundary that also happens to be the commercially winning wedge today (outreach, recalls, scheduling, HRA/AWV, chronic care, trial logistics). The Aug 2026 shift from point agents to 30+ outcome-focused Agentic Orchestrators is the bet that moves it from per-task SaaS per-minute economics to population-level outcomes contracting, exactly when payers (Star Ratings, HEDIS, readmissions) are hungry for measured ROI. Risks: healthcare procurement cycles are long, per-interaction pricing caps the ceiling unless orchestrators hit outcome pricing, and foundation-model commoditization plus well-funded entrants (Microsoft/Nuance, ambient-AI players, payor-built agents) could erode the wedge. Overall, a rare 3x (Seed 2023 → $1.64B Jan 2025 → $3.5B Nov 2025) with genuine clinical proof.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki