Healthcare has a pattern with new technology: overpromise a replacement, underdeliver on the integration, and leave the actual clinical relationship worse off than before it arrived.
Myra Ahmad’s read on artificial intelligence, laid out on the Compound Wisdom podcast in April 2026, is built to avoid repeating that pattern. AI is genuinely powerful for the backend of healthcare, operations, workflows, efficiency, and not yet ready, and maybe never suited, to replace the judgment a physician exercises with an actual patient in front of them.
For a founder who built Mochi Health around the idea that patients fall out of care in the gaps between systems, the distinction isn’t academic.
Where AI gets deployed inside that architecture, and where it explicitly doesn’t, determines whether the technology closes those gaps or quietly creates new ones.
Where AI Actually Earns Its Place
Ahmad has long argued that one of healthcare’s core failures is structural fragmentation: “most providers are getting labs from separate systems that never talk to each other,” as she’s put it, with medications, the provider relationship, and the treatment plan sitting in disconnected silos.
That is precisely the kind of problem AI is well suited to help solve. Pattern-matching across large volumes of disconnected data, flagging a lab trend before it becomes a crisis, or surfacing a drug interaction buried in a patient’s history, is exactly the operational work that benefits from a system that never gets tired and never misses a row in a spreadsheet.
The same logic applies to the unglamorous parts of running a healthcare marketplace: routing patients to the right provider or pharmacy, catching a scheduling conflict, or streamlining the clinical oversight work that has to happen behind every prescription. None of that requires clinical judgement about an individual patient’s care.
It requires speed and consistency at scale, which is where AI’s real advantage over a human system currently sits, and where the return on the technology is the most immediate and the least controversial.
Where Physician Judgement Still Has To Lead
The clearest argument against letting AI make clinical calls comes directly from Ahmad’s own position on dosing. In her MedCity News op-ed, she argued that individualized GLP-1 dosing “demands greater medical supervision, not less,” precisely because personalization introduces variables that a rigid protocol doesn’t have to account for.
An algorithm trained to optimize for the average patient reproduces the exact failure she has spent years criticizing in clinical protocols generally: models built around an average trial population that doesn’t represent the person actually being treated.
Handing that decision to AI without a physician in the loop automates bias at scale.
There’s also the trust question, which Ahmad has framed as something AI cannot substitute for. A patient disclosing a side effect they’re embarrassed about, describing a symptom they can’t quite articulate, or explaining why they actually stopped taking a medication is context a physician can read and act on in ways a model trained on structured data cannot.
The most consequential clinical decisions tend to happen exactly where the data is incomplete and the judgment call is the point, not where a pattern is already obvious enough for software to catch it.
That’s a meaningful distinction from how the debate is often framed. The question usually asked is whether AI is accurate enough to be trusted with a clinical decision.
Ahmad’s framing asks a different question: whether the decision in front of a physician is the kind that a model, however accurate on average, was ever built to make for an individual.
The Compound Wisdom Argument, Applied To Mochi
That division of labor is already how Mochi Health is built. The company’s marketplace software exists to make the operational side of care faster and more connected, matching patients with providers and pharmacies, and surfacing the data a provider needs at the point of a decision.
What it doesn’t do is put a layer between the provider and the choice they make for a specific patient. Providers on Mochi Health, in Ahmad’s framing, practice without the interference of a system dictating outcomes to them, and that same principle is what keeps AI positioned as infrastructure rather than as a decision-maker.
Setting The Terms Instead Of Reacting To Them
Most of the industry is still oscillating between AI hype and AI panic, either promising an algorithm will replace a doctor’s judgment or reflexively rejecting the technology altogether. Ahmad’s position is a specific, defensible line: let the technology do what it’s actually good at, and keep the physician firmly in charge of everything else.
That’s not a hedge against AI. It’s the argument for building with it correctly from the start.