Training clinical AI agents the way medicine trains doctors - through residency, simulation, and a safety bar earned before the first real patient.
Healthcare in the United States rarely fails for lack of ambition. It fails for lack of hours. There are more patients who need intake, triage, reminders, follow-up and answers than there are clinicians with time to give them. Amigo, a New York company founded in 2023, is building software to absorb that overflow - not with a chatbot bolted onto a website, but with AI agents trained to work inside real clinical operations.
The company builds a platform that lets healthcare organizations design, train, deploy and manage patient-facing agents. Those agents handle intake and triage, care navigation, and around-the-clock patient support, and they plug into the electronic health records that hospitals already run on - Epic, Oracle Health and Athenahealth among them.
The name is not accidental. Amigo means friend, and the agents speak more than 100 languages - a practical answer to the fact that navigating care is hardest for the people least equipped to fight through it.
That framing is personal for co-founder and CEO Ali Khokhar. He was eight when his mother was diagnosed with breast cancer, and he watched his family struggle with the logistics of her care before she died when he was fourteen. Amigo is, in part, the tool he wishes his family had had.
“We train our agents like doctors because mistakes can cost lives in healthcare.”
What sets Amigo apart from the crowded field of healthcare AI is less the model than the training regimen. Before an agent ever speaks to a real patient, it completes what the company calls a digital residency: it practices across millions of simulated patient scenarios built around a specific practice's actual population.
Those simulations include virtual patient personas, edge cases and supervised evaluations. Amigo says an agent must reach a 100% safety pass rate in testing before it is allowed into production. The company reports zero safety incidents across more than three million real encounters.
A second, quieter layer sits underneath: a data foundation that ingests messy, unreliable healthcare data from every source, cleans and unifies it, and feeds the agent engine, EHR connectors and analytics. Fix the foundation first, then automate on top of it.
Figures reported by Amigo, March 2026.
Tools to design, train, deploy and manage patient-facing agents for intake, triage, care navigation and 24/7 support.
Simulation-based training across millions of scenarios tailored to a practice's population, with a 100% safety pass rate before deployment.
An ingestion layer that cleans and unifies messy healthcare data and serves it to agents, connectors, operator workflows and analytics.
Connections to Epic, Oracle Health, Cerner and Athenahealth so agents operate inside existing systems of record - not beside them.
Amigo sells to healthcare organizations - digital-health companies, clinics and health systems. Publicly referenced customers include Eucalyptus, Hazel Health, Wisp, Jasper Health, Diverge Health and The Care Clinic, among more than 30 organizations, with Stanford Health Care named in its orbit through advisory ties.
The market it plays in - clinical AI agents and patient engagement - is competitive, with players such as Hippocratic AI and Infinitus pursuing overlapping ground. Amigo's differentiator is not that it deploys agents but the discipline it puts in front of deployment: simulation, supervised evaluation, and a refusal to ship below a safety bar.
The business model
Amigo runs a B2B SaaS model with two revenue streams: a monthly platform fee covering agent engineering, deployment strategy and training infrastructure, plus usage-based fees as agents handle patient interactions. The more work the agents do, the more the model scales - which is why the safety regimen matters commercially as well as clinically.
“Amigo's approach of training agents with the same rigor we expect of clinicians means they can operate at the standard I've seen at Stanford, Columbia, and MD Anderson.”
| Round | Amount | Date | Lead / Investors |
|---|---|---|---|
| Seed | Part of $17M total | 2024 | General Catalyst, GSV Ventures |
| Series A | $11M | Mar 2026 | Madrona (lead); Optum Ventures, General Catalyst, GSV Ventures |
The Series A was announced in March 2026, bringing cumulative funding to roughly $17M.
Ali Khokhar and John Xing start Amigo in New York to build safe, patient-facing clinical AI agents.
Seed funding co-led by General Catalyst and GSV Ventures; the simulation-based training model takes shape.
Agents go live with Epic, Oracle Health and Athenahealth and begin accumulating millions of encounters.
Madrona leads an $11M round; Dr. Jay Shah joins as Chief Medical Advisor; encounters cross 3 million with zero safety incidents.
It provides a platform for healthcare organizations to build, train and deploy patient-facing AI agents for intake, triage, care navigation and 24/7 support, integrated with their EHR systems.
Ali Khokhar (Co-founder & CEO) and John Xing (Co-founder).
Each agent completes a “digital residency,” training across millions of simulated scenarios and reaching a 100% safety pass rate before touching real patients. Amigo reports zero safety incidents across 3M+ encounters.
About $17M total, including an $11M Series A announced in March 2026 led by Madrona, with Optum Ventures, General Catalyst and GSV Ventures participating.
Major EHRs including Epic, Oracle Health, Cerner and Athenahealth. Agents operate in 100+ languages.