AI tools are increasingly present in the exam room — but a handful of startups want AI to replace doctors.
getty
Today, most clinical AI companies concentrate on tools designed to expedite doctors’ work, aiming to reduce burnout and assist in more informed decision-making. For instance, AI scribes like Abridge and Cleo cut down documentation time, while medical-focused AIs such as OpenEvidence support doctors in making better decisions.
However, a smaller number of startups are venturing beyond assisting doctors and are exploring ways to replace them altogether. Here are three companies leading this charge.
Doctronic: the AI doctor with prescribing power
Doctronic, based in New York, has developed a HIPAA-compliant platform marketed as a 24/7 personal AI doctor capable of making its own clinical decisions. In March, the company secured $40 million in a Series B funding round, co-led by Abstract and Lightspeed Venture Partners, reaching a total funding of $65 million.
This startup is operational in Utah, where it legally renews prescriptions for chronic condition patients without needing a doctor’s approval. This marks the first instance of a state allowing an AI platform to independently handle routine prescriptions.
After six months, data revealed that the AI managed to approve 72% of prescription renewals for 192 drugs, including those for hypertension, diabetes, and depression, without human intervention. The system escalated to a human physician in 28% of cases, typically when additional steps like lab tests were required or complications were flagged. Independent physician reviewers agreed with the AI’s decision to escalate in 69% of these instances.
In cases where the AI approved a renewal without escalation, physicians concurred with its decision 91% of the time, though 3% of the time, they disagreed entirely with the AI’s judgment.
Utah’s Office of AI Policy reports no serious safety incidents have been reported. Nevertheless, the Utah Medical Licensing Board and the American Medical Association have both expressed concerns about accountability, and researchers have identified security vulnerabilities that regulators are addressing.
Certuma: an AI replacement for urgent care
Founded by serial entrepreneur Martin Varsavsky, Certuma focuses on common conditions that frequently occupy urgent care facilities. Based in Austin, the company has raised $10 million in seed funding, led by Joe Lonsdale’s 8VC, with a valuation of $60 million.
Certuma aims to become the first AI doctor formally approved by the FDA. It initially targets 25 low-risk, high-frequency conditions, such as UTIs, strep throat, and sinusitis, chosen for their well-established and low-variance clinical decision trees.
The company is pursuing a two-track strategy: a more rigorous FDA approval path in the U.S. and a faster international track in Argentina, Varsavsky’s home country. There, aided by supportive government deregulation, Certuma has already collaborated with regulators to provide diagnostic and prescribing advice via a consumer-facing chatbot.
The long-term goal is to integrate the AI with autonomous robotic booths for tasks like imaging or blood draws.
Ada Health: AI triage without a waiting room
Ada Health takes a different approach, offering an AI health assistant that helps users understand symptoms and decide on next steps — often without consulting a doctor.
The Berlin-based company has raised $242 million, with its largest funding being a $120 million Series B round supported by investors like Inteligo Bank, Farallon Capital, and Red River West.
Ada Health claims 13 million users globally, with over 32 million symptom assessments completed.
A peer-reviewed BMJ study led by Ada Health compared eight symptom-checker apps on their coverage, accuracy, and safety, with Ada’s tool ranking highest. Though more of an “AI front door” than a full AI doctor, the end result is fewer visits requiring a physician.
Bigger AI players are building copilots, not replacements
In contrast, major AI companies are developing tools to assist rather than replace doctors. Anthropic introduced Claude for Healthcare in January 2026, at the J.P. Morgan Healthcare Conference, as a tool designed to aid physicians. Most large health systems have similarly positioned AI for managing documentation, triage support, and administrative tasks, allowing doctors more time for complex decisions.
However, the boundaries are becoming blurred. Revere Health has already initiated AI-related workforce reductions in the U.S. healthcare sector, impacting medical coding, billing, and documentation roles, but not doctors. About 200 jobs, or 7% of the workforce, were cut. While physicians remain unaffected, the quiet erosion of roles around them may indicate future pressures.
In a recent Sermo poll, 58% of physicians expressed concerns that AI could reduce their role or render doctors obsolete.
Here’s the regulatory patchwork that’s making this possible
Replacing doctors with AI is not straightforward. Two regulatory areas must evolve: federal device approval and state medical licensing. Both are currently being tested by various companies.
Federally, the FDA has cleared over 1,000 AI/ML-enabled medical devices, mostly through the 510(k) pathway, which requires a device to be “substantially equivalent” to an existing market product. This pathway doesn’t support autonomous treatment decisions.
The De Novo pathway, for novel low-to-moderate-risk devices, has set a precedent with LumineticsCore (formerly IDx-DR), authorized in 2018 as the first fully autonomous AI diagnostic system for diabetic retinopathy detection without physician input.
Certuma aims to replicate this model for its 25 target conditions. However, no autonomous AI prescribing service has yet been cleared via any FDA pathway.
On the state level, the approach differs and could have more immediate implications. Utah didn’t use a licensing exception for Doctronic — the company isn’t licensed to “practice medicine.” Instead, under an AI regulatory sandbox law passed in 2024, Utah’s Office of AI Policy signed a non-enforcement agreement, allowing Doctronic to operate without prosecution, provided it adheres to specific safety and privacy requirements. This covers only prescription renewals, not new diagnoses or first-time prescriptions.
Yet, this carve-out faces opposition from new legislation in places like Oregon and Delaware, which prohibit any “nonhuman entity” from using licensed clinical titles or practicing.
By mid-2026, about 37 states had introduced or passed AI-in-healthcare legislation, with common themes requiring licensed human clinicians to make final decisions on medical necessity. Utah’s approach is exceptional and is already receiving criticism from its Medical Licensing Board and the AMA.
Doctronic and Certuma argue that if algorithms can manage the routine 80% of cases, human physicians can focus on the more complex 20%, providing patients with quicker access.
Conversely, the concern is that medicine involves more than pattern recognition. A sore throat could be strep or an early sign of something more serious, which an algorithm might overlook but a skilled clinician could detect.
The resolution of the Utah prescription initiative, FDA submissions, and unresolved malpractice questions will ultimately determine the direction of this story.
The prevailing belief is that AI will transform medicine rather than replace its practitioners. Nevertheless, for the first time, venture-backed companies, state governments, and regulators are actively exploring the implications if that belief proves incorrect.
