How to Sell AI to Hospitals: The Buying Committee Playbook

Selling AI to hospitals is committee selling with a compliance gauntlet welded on. No single person can say yes, but five or six people can each say no, and the deal closes only when all of them stop blocking at the same time. I sold to hospital CIOs and clinical leadership from 2018 to 2026 at Verto Health, where I joined as employee #2 and we grew to 80 people and 100+ deployments across Canadian and US health systems. That structure held on essentially every deal I touched, from single-program pilots to platforms that ended up orchestrating 10 million+ patient journeys.

The good news for anyone trying to sell AI to hospitals is that the committee is predictable. The same chairs show up at almost every hospital, they ask roughly the same questions, and you can prepare for each one before you walk into the building.

The Five-Chair Map: know who can kill your deal#

The Five-Chair Map is a one-page exercise that names the five people who can each veto your hospital deal, plus the single question each of them needs answered before they stop blocking. I never called it that inside Verto (it lived as a habit, not a slide), but it is the thing I rebuilt at the start of every serious pursuit.

The steps:

1. Put a real name in every chair before your second meeting. "The privacy office" is not a name. If you cannot name the person, that chair is voting no by default.
2. Write down the veto question each chair is silently asking. They rarely ask it out loud during the demo. Defaults are below.
3. Build one artifact per chair, not one deck for everyone. A workflow walkthrough, an architecture diagram, a data-flow map, a budget memo. Five chairs, five artifacts.
4. Ask your champion which chair is coldest, then go there first. Cold chairs get colder while you celebrate the warm ones.
5. Re-run the map every quarter. Hospital leadership rotates constantly, and a deal this long will outlive at least one org chart.

Here are the chairs.

Chair 1: the clinical champion#

Usually a physician lead, program director, or department chief. Their silent question: "Will this make my clinicians' day shorter or longer?" They do not care about model accuracy in the abstract. They care about clicks, alert fatigue, and whether the nurses will quietly revolt. I ran product across 14 client working groups, and this was the chair where meetings were actually won or lost: the moment a clinician says "that's not how intake works here," your credibility drops to zero. Demo their workflow, in their vocabulary, or do not demo.

Chair 2: the CIO#

The CIO owns the architecture and the risk register, and your product lands on both. Their questions are concrete: where does this sit relative to the EHR, does it speak FHIR, how does patient identity resolve (the enterprise master patient index, or EMPI, matters more than most vendors expect), how does SSO work, and what is the exit plan if this fails. I ended up doing the architecture work myself at Verto (FHIR, EMPI, SMART on FHIR), which taught me the real test: the CIO is not evaluating your product. They are evaluating whether you will become a two-year cleanup project on their watch.

Chair 3: the privacy and security officer#

Their question: "Show me exactly where patient data goes, who can touch it, and what happens when the system is wrong." A privacy impact assessment and a security review are coming whether you like it or not. The strong move is to arrive with the answers pre-written, which I cover in the next section.

Chair 4: finance and procurement#

Their question: "Which budget line does this live on after the pilot money runs out?" Pilots often ride on innovation funds or one-time grants. Operating budget is a different pool with a different approval chain, and if nobody can name the line your renewal comes from, you do not have a customer. You have a pilot with an expiry date.

Chair 5: the workflow owner#

The nurse manager, unit clerk, or registration lead. Lowest title in the room, most practical veto in the building. Their question: "How many clicks are you adding to my worst day?" This chair never escalates a formal objection. Usage just flatlines, and six months later the committee reads the adoption report and concludes the product failed.

The compliance gauntlet is the sales process, not paperwork after it#

Every hospital deal runs through some stack of privacy impact assessments, security and threat-risk reviews, procurement rules, and increasingly a dedicated AI or ethics review on top. Vendors tend to treat this as friction that starts after the win. That framing loses. The gauntlet is how the hospital decides whether you win.

The practical move is a living compliance package: data-flow diagrams, hosting jurisdiction, subprocessor list, model behavior and failure modes, breach protocol, audit logging. Hand it over before anyone asks. At Verto, architecture answers were sales answers; the multimillion-dollar demos I built worked because they could survive a security architect in the room, not because they were pretty. When your buyers are hospital CIOs and clinical leadership, competence in the gauntlet is the differentiation.

Also, calibrate your calendar. Six to eighteen months from first meeting to signed enterprise contract is normal, not a sign your deal is broken. Founders who plan for a 90-day cycle run out of patience (or runway) right when the committee is finally warming up.

Pilots are where healthcare AI goes to die, unless you design the exit#

The pilot-to-enterprise gap deserves its own article, but the sales-side rules fit here:

1. Write the expansion criteria into the pilot contract itself. Agree on the numbers up front: what adoption, throughput, or time-saved threshold triggers the enterprise conversation. A pilot without a defined exit is a science project with an invoice.
2. Fund the pilot as close to operating budget as you can. Innovation-fund pilots evaporate when the fund does. A pilot the department pays for, even partially, has an owner with skin in it.
3. Staff the pilot like production. A pilot that gets your intern gets their intern, and their intern cannot push an enterprise deal through five chairs.

Expansion is also where the business model actually works. Verto's ARR tripled two years running, and you do not get that kind of compounding in hospitals from new logos alone. The motion that compounds is land, prove, expand inside the same health system, where the compliance work is already done and the chairs already know you.

What the RFP actually tests#

I wrote a lot of RFP responses over eight years and won enough to see the pattern. The RFP is not testing features. It tests four things:

• Whether you will exist in three years. Financial disclosures, team depth, reference customers who pick up the phone.
• Whether you survive the security and privacy annexes. These sections are scored by people who never saw your demo.
• Whether you understood the workflow or just the keyword list. Generic answers about "improving patient outcomes" score like the generic answers they are.
• Whether you shaped the requirements before the RFP posted. This is the quiet one. The best position is when the requirements read like your product because you spent six months educating the committee before procurement opened. If the first time you see the RFP is the day it posts, there is a decent chance you are column fodder in someone else's evaluation matrix.

What kills deals#

The recurring killers, in rough order of how often I watched them happen:

• Champion turnover. Your map goes stale, and the new director inherits zero of the trust you built. This is why step 5 of the Five-Chair Map exists.
• Letting "AI" do your positioning. Leading with the technology can trigger review cycles the workflow framing never would. Sell the outcome (fewer no-shows, faster discharge, less duplicate data entry) and disclose the AI fully in the compliance package, where it belongs. This is honesty plus sequencing, not concealment.
• No budget line after the pilot. Chair 4, unanswered.
• Integration surprise discovered late. An identity-matching or interface problem that surfaces in month nine restarts the CIO's risk math from zero.
• Artifact mismatch. Showing clinicians an architecture diagram and the CIO an outcomes story. Each chair got an answer, just not theirs.

None of these are exotic. That is sort of the point: hospital deals rarely die from a competitor being better. They die from an unmapped chair.

Where to go from here#

If your specific problem is a pilot that will not convert, start with why healthcare AI pilots die. And if you are a founder with a healthcare AI product, a pilot in flight, and a committee you cannot fully name yet, that gap is exactly the kind of thing I work on hands-on. The engagement shapes are on the engagements page. Mapping the five chairs on a live deal takes about an hour and usually changes the next three months of your calendar.