General AI Is Built to Be Right About Everything. Daily Command’s Semmelweis Is Built to Be Right About the Next Prescription

This piece was written by Harshit Jain, MD, Founder & Global CEO, Doceree 

Every platform in healthcare marketing now claims AI. The question that decides outcomes is different — what was the intelligence underneath built to be right about? Here is our answer.

“Powered by AI” now appears on most platforms in healthcare marketing, and it means less every time it appears. True or not, the label answers nothing. What decides outcomes is what the intelligence underneath was built to be right about. 

General-purpose AI was built to be right about everything: the open web, the average case, the plausible answer. That is its limitation in pharma, where the decisions that matter are not average and the cost of plausible-but-wrong is regulatory. So we built Semmelweis, the model powering Daily Command, to be right about exactly one thing: the next prescription. 

The distinction that decides everything 

Semmelweis distinguishes between a physician researching a therapy and one about to prescribe it, a distinction no open-web model can make, because nothing on the open web contains it. That distinction is the difference between a well-written message and a well-timed one; between reach and relevance. 

It exists because of what we chose to ground the model in. Where horizontal AI understands the internet, Semmelweis reasons over deterministic clinical context: the ESPYIAN identity graph of more than 6 million verified HCPs, live Clinical Intent Signals detecting prescribing intent as it forms and 185+ direct EHR integrations. It knows where a physician stands in a treatment decision instead of guessing from demographics. 

The foundations of Semmelweis 

Claims like ours hold or collapse at the foundations, so here is exactly what we did. We did not adapt a general-purpose model for healthcare, and we did not build a foundation model from scratch. We built on frontier models for language and reasoning; competing with them was never the point. What sets Semmelweis apart is the proprietary architecture built around them. We knew from the start that defensibility would never live in model weights, which are licensed and commoditize by the quarter. It lives in the architecture around the weights, and it strengthens with use. 

We built three capabilities into that architecture that no general-purpose model can replicate. The first is that clinical grounding: deterministic clinical context the open web does not contain. The second is a compliance-bounded action space: The model never touches patient data, operates solely on de-identified, HCP-level context and recommends only actions that already sit within medical, legal and regulatory guidelines. Every output is advisory, human-approved, auditable and MLR-ready by design. The third is a closed measurement loop: Each outcome the model observes informs its next recommendation, so precision improves with every campaign. In pharma, the second capability is not a constraint on the first. It is what makes the intelligence usable at all. We treat privacy and regulation not as limitations but as the operating surface itself. 

A name that carries a lesson 

I trained as a physician, so the story of Ignaz Semmelweis was not new to me: the doctor who discovered in the 1840s that handwashing saved lives. He was rejected, ridiculed and driven out by the medical establishment for saying so, and vindicated only after his death. Semmelweis saw the signal before the industry was ready to believe it, which is what our model is built to do. 

We could have shipped a feature and moved on. We named it instead, because the lesson matters. Our model is built to see the signal early, and to act on it only within the guardrails this industry rightly demands. It is the first model we have named. It will not be the last. 

What it means inside Daily Command 

For brand teams, the model is invisible; its effects are not. Daily Command, the first commercial operating system for pharma, launched July 14. Powering it is Semmelweis, sitting above the DSPs, networks and channels a brand already uses, orchestrating across them rather than working inside one. Recommendations arrive timed to the decision rather than the media calendar, and every outcome feeds the next one. 

My answer to the AI question in healthcare marketing is not “more.” It is intelligence built for the work.

Semmelweis is live at doceree.com/semmelweis.