Selling AI to Hospitals for Eight Years: The Verto Story

I spent eight years at Verto Health selling AI into hospitals. I joined as employee #2 and helped scale the company to 80 people. Along the way: 100+ deployments across Canadian and US health systems, and something north of 10 million patient journeys orchestrated on the platform. This is the case study version of those years. I wrote it the way I would explain it to a founder deciding what that experience is actually worth to them.

The real job of employee #2 is a queue#

When there are two of you, the job title is a formality. The real job is a queue of things the company does not have someone for yet, sorted by whatever is currently on fire. Early on the queue is short and obvious: design the screens, write the requirements, make the demo work. As the company grows toward 80 people, the queue changes character. Every new hire takes over a lane, and your job becomes the lanes that do not exist yet.

Over eight years that added up to six or seven distinct roles, depending on how you split the last one. Not shadowed. Owned, with deliverables and consequences attached. I want to go through them one at a time. "Wore many hats" is the least useful phrase in hiring; the specifics are the entire point.

What "doing whatever the company did not have someone for" meant#

Product design#

I produced more than 10,000 hi-fi screens over those years. Hospital software gets used by a nurse at hour eleven of a twelve-hour shift. It also gets used by a registration clerk with a line of patients watching them type. Clinical users do not forgive ambiguity. Design in that world means removing every decision the screen could make for the user instead.

Product management#

I ran 14 client working groups as a PM. A hospital working group is its own genre of meeting: clinicians, IT, privacy officers, sometimes more, most holding a veto and none holding the final say. You learn to write decisions down and name owners, because the alternative is a project that drifts for a fiscal year.

Software architecture#

FHIR, EMPI, SMART on FHIR. In plain English: FHIR is the standard hospitals use to exchange health data, and an EMPI is the index that decides which records across different databases belong to the same patient. Health systems buy compatibility with the systems they already own; the software itself comes second. So the architecture work was mostly about fit. Patient identity across databases that disagree on who the patient is, and standards work so integrations survived contact with real hospital IT.

Data science#

Digital twins, meaning live data models that mirror each patient's real-world journey, tracked more than 9.5 million encounters. NLP and OCR pipelines read documents that still arrive as faxes. Ontology knowledge graphs made medical coding systems talk to each other. This thread eventually turned into a patent, which I will get to below.

Sales engineering#

Multimillion-dollar demos and RFP wins. The stages included HLTH in Vegas, Expo.Health in Boston, MaRS, BMZ, and Plug and Play Creasphere in Munich, where the Verto x NEOM partnership won in 2024. A demo for a hospital CIO is a strange artifact. It has to look finished, survive hostile questions from clinical leadership, and still honestly represent a product that is being built underneath it.

Product marketing and AI transformation#

The final title was Director of Product Marketing and AI Transformation, with 30+ GTM strategies to show for it. In practice that meant converting what the other roles taught me into positioning, launch plans, and credible answers to "why you and not the incumbent."

The pattern across all of them was the same. The company needed the function before it could afford the specialist, so I became the function until the specialist arrived, then handed it off and moved to the next gap. Building a function you fully intend to give away turned out to be the most transferable skill of the whole run.

The COVID-scale moment#

In 2020 and 2021, health systems had to vaccinate entire populations with software assembled in weeks. At peak, the Verto platform was helping run vaccination programs across provincial health systems. The integration with McMaster Health Labs ran the world's largest COVID study in 2020.

I will not pretend that period was a controlled experiment in anything; it was scale arriving faster than plans could be written. But it settled the question every founder eventually faces about their own product: what happens when real volume hits? We found out in public, with provincial health systems watching. The operational habits from that stretch never left me: ship daily, and assume the requirements change tomorrow.

The quieter proof point from those years was the Bayshore patient-transition platform, which won multiple awards. Less dramatic than vaccine volume, but closer to the everyday job of the product: moving patients between care settings without dropping anyone.

The patent nobody plans to write#

In January 2025 I co-authored patent WO2025147762A1: dynamic clustering of medical coding systems for patient journey reconstruction. Behind the dry title is a stubborn problem. Hospitals encode the same clinical event in different coding systems, so reconstructing one patient's journey across institutions is genuinely hard. The patent grew out of years of the data science work above, not out of a research mandate. I mention it because it is the kind of artifact that only appears when someone stays with one domain problem for the better part of a decade.

Tripled ARR, twice, and what hospital buyers actually purchase#

The commercial result: ARR tripled two years running. The buyers were hospital CIOs and clinical leadership, which is roughly the hardest B2B audience I know of. Procurement cycles run long. Decisions run through committees, and the privacy and safety obligations are real rather than performative. These buyers also carry a long memory of vendors who overpromised.

The core lesson from selling to them is short. Hospitals do not buy AI; they buy risk reduction that happens to have AI inside it. The demo matters more than the deck, references matter more than the demo, and nothing matters more than surviving your first deployment. I wrote the tactical version up separately in how to sell AI to hospitals, if you want the playbook rather than the memoir.

What a founder hires from eight years of this#

That chapter closed for me in 2026, and the question founders ask now is what it converts into. The honest answer is three shapes, which map directly to how I structure engagements:

Fractional GTM operator. Positioning, demand, and the marketing function itself, done hands-on rather than advised from a distance. The Verto version of this was 30+ GTM strategies aimed at the least forgiving buyers in software. The operating framework I now use for the marketing side is public, as the Agentic Marketing OS.

Product and build. Brief to shipped product, with six weeks as a working number. The 10,000 screens and the architecture years are why I can go from whiteboard to production without a ten-person team in between.

AI operations advisory. Where agents fit and where they do not, delivered as working pilots rather than slide decks. The healthcare years are the source of the skepticism here. I have watched AI succeed under some of the least forgiving conditions available, and I have watched what kills it.

One caveat I give everyone: this experience transfers best to founders who decide directly and move fast. If your organization needs three layers of approval before anything ships, I am the wrong hire, and no amount of hospital scar tissue changes that.

Where to go from here#

If you are still weighing what this kind of hire even is, what a fractional operator does is the right next read. If one of the three shapes already sounds like your gap, the engagements page above explains how the days and weeks work. A call makes sense once you know which shape you are hiring for; until then, the reading is cheaper.