Matt Pavelle

The through-line

Everything Matt Pavelle says returns to one claim: the information-gathering half of medicine is now machine work, and once you accept that, both access and physician burnout become tractable problems at the same time. Doctronic is built as a direct-to-consumer AI doctor with a telehealth practice behind it, free and anonymous by default, with human doctors available around the clock in all 50 states for about $39 a visit or through insurance [1]. He is blunt that this is not AI as garnish on an existing model: "what we're doing could not be done without AI" [1]. The scale he cites is the argument for feasibility rather than novelty, 17 million patient interactions and 150,000 people a week using the AI [1], with the AI positioned as the new first step in patient care rather than a curiosity [4].

The position has been getting steadily more assertive rather than changing direction. It began as an AI that prepares the ground for a physician, and has moved toward AI holding formal standing of its own, with Doctronic described as the first AI in the world legally licensed to practise medicine, through Utah's AI Learning Lab regulatory sandbox in December 2025 [7][2].

On what the AI actually does before a doctor ever sees you

The product is a conversation, not a symptom checker. There is no login required, the website is anonymous and free by default, and the AI opens by asking what is going on rather than demanding structured input [1]. A typical conversation runs 20 minutes, and 80% of people who start one finish it, which he attributes partly to the fact that it happens on the patient's schedule: "start the conversation, go cook dinner, go take your kids to soccer, we don't care, then continue the conversation" [1]. The output is deliberately shaped like clinical reasoning rather than an answer, "the four most likely diagnoses and here are four treatment plans" presented as a differential, "it's probably this, but it could be this, this, this" [1]. Two artefacts come out of it: an explanation written so a patient can understand it, and a note the patient can carry to any physician, including one who has nothing to do with Doctronic [1]. The same architecture is what he describes as gathering history, following clinical guidelines and producing full treatment reasoning [4]. Behind it sits an agentic system with multiple pathways, of which diagnosis is only one [1].

On accuracy as the precondition for everything else

The number he leads with is a 99.2% treatment alignment between what the AI recommends and what the Doctronic physicians who then see those patients agree on [1][6][8]. He frames the engineering achievement as taking existing foundational models and making them accurate, grounded in real medical evidence [1]. The clinical metaphor he uses is his doctors' own: the AI "does the work of the best chief resident in the world" [1], meaning it gathers everything, proposes likely diagnoses and matching treatment plans, and hands the whole package over. The alignment rate is what licenses the rest of the argument, and it is what the industry has picked up on in treating Doctronic as a trusted AI doctor rather than a demo [6][8].

On giving doctors back the part of the job they wanted

His sharpest material is about physician burnout, and it is personal: his wife is a doctor, and his mother-in-law sees patients Monday through Friday and then spends Sunday catching up on charting [1]. He calls the notes problem pajama time and dismisses the whole ritual as "just so archaic and old" [1]. Doctronic's answer is to invert the sequence. The AI writes the doctor's note ahead of the visit, so that if the physician agrees with the top-ranked diagnosis and treatment, "their charting is done" and they are "more often just deleting a couple words from the note. They're never writing it" [1]. The clinical consequence matters more to him than the time saved: instead of spending the visit asking about medications and family history, "our doctor gets to just look at all that information that is passed to them... and then have a meaningful conversation with the patient. So the care is a lot better and it really is care" [1]. His underlying belief about motivation is simple and he repeats it in different forms: "Doctors love treating patients. That's why they're doctors" [1].

On what AI should take over next, and what it should never touch

He draws the line by asking what doctors went to medical school for. His own statin renewal is the worked example: an annual blood test and an in-person cardiology visit that he says AI will never take, against a quarterly prescription renewal that involves days of asynchronous back-and-forth with a PA, which "within 20 years an AI can probably do that safely" [1]. Starting a prescription remains "a doctor thing for sure" [1]. He generalises this into a view of incremental redesign rather than replacement: "there are lots of little tweaks in the system like that where AI can be super helpful years from now," specifically the things doctors currently do that they do not want to do [1]. The urgency behind the refill case is not abstract for him, since he frames failure to access refills for chronic conditions people already know they have as a cause of 150,000 American deaths a year [2].

Asked what success would look like in five years, he does not name a product. He names a permission: "Can we get all of the payers okay with the AI writing all of the insurance paperwork and a doctor just reviews it?" [1]. The reasoning is the same as with notes, that review is far cheaper than authorship, "it's a lot faster to review than to write. Can we normalize that?" [1].

On access, cost and regulatory legitimacy

Access is the framing he puts around the whole business: "we are broadening access, right, changing the way people do access healthcare" [1], with millions of patients using the AI and close to hundreds of thousands seeing physicians [1]. The pricing and the anonymity are part of that argument rather than growth tactics, since the AI tier is always free and the visit is deliberately cheap and available 24/7 [1]. He has also argued the case for AI reshaping access to medicine publicly alongside co-founder Adam Oskowitz [3]. The regulatory track runs in parallel: rather than operating in ambiguity, Doctronic went through Utah's AI Learning Lab regulatory sandbox to become the first AI legally licensed to practise medicine, in December 2025 [7][2]. The consistent instinct across all of it is to make the AI's role explicit and sanctioned, whether by payers, regulators or the doctor reading the note.

Takeaways

• The AI is positioned as the first step in care, not a novelty layer: it gathers history, follows clinical guidelines and produces full treatment reasoning before any human is involved [4][1].
• Doctronic reports 99.2% alignment between the AI's proposed treatment and what its own physicians conclude after seeing the patient [1][6][8].
• Notes are written before the visit, so physicians typically delete a few words rather than compose anything, which is Pavelle's direct answer to pajama-time charting [1].
• Free and anonymous by default with no login, a 20-minute conversation completed by 80% of people who start it, and a $39 or insurance-covered visit if the patient wants one [1].
• The line he draws on automation: renewing an existing statin prescription is safely automatable within 20 years, starting that prescription and the annual in-person cardiology visit are not [1].
• His five-year goal is regulatory and administrative, getting payers to accept AI-written insurance paperwork with a doctor reviewing rather than writing it [1].
• Doctronic went through Utah's AI Learning Lab regulatory sandbox to become the first AI legally licensed to practise medicine, in December 2025 [7][2].