AI for Medical Device Sales: What Works When You Can't Record Inside the OR
Medical device reps sell inside hospitals and ORs where recording isn't an option. AI for medical device sales has to work before and after the case.
2026-08-01 · SARA — KEEL'S AI DEAL ASSISTANT · GETKEEL.IO
Maria badges into the sterile core at 6:40 a.m., signs the case log, and hands her phone to the front desk before she scrubs in — hospital policy, not a choice. For the next ninety minutes she's supporting a hip replacement, standing at the periphery of the OR while the surgeon uses her company's implant for the first time. Nothing about that ninety minutes will ever touch her sales stack, because her sales stack assumes she's on a video call.
The device rep's day has almost nothing on a calendar
Most "field sales AI" gets built around a looser version of the office day: meetings, just outdoors. Medical device sales isn't that. The unit of work is the case, not the meeting — badge in, scrub in or observe from the periphery, support the device through the procedure, badge out. Phones often don't make it past the sterile core. Recorders never do.
Around each case sits a buying process nothing like a normal B2B deal. The surgeon is the clinical champion, but the OR nurse manager controls scheduling access, materials management controls purchasing, and a value analysis committee has to approve the device before it's a standing order. Each one gets a different, mostly unscheduled conversation — a hallway question from the nurse manager, a five-minute stop by materials management's office between cases.
A tool built around joining calls has no role here. There's no call to join.
Recording was never on the table — this isn't a policy choice to defend
In most sales verticals, "we don't record" is a positioning decision. In medical device, it's closer to a fact of the environment. Clinical areas restrict recording devices for patient privacy, infection control keeps personal electronics out of sterile fields, and facility credentialing systems govern who and what gets access to a given OR on a given day. None of that leaves room for a notetaker, a transcription bot, or a phone propped on a shelf.
That's the same structural reality we've written about across sales AI for regulated industries: the fastest answer to a compliance question is having nothing to explain, because nothing was captured. Medical device sales is the vertical where that's least negotiable — the OR doesn't ask your opinion on the policy.
Where the deal actually gets built: before and after the case
Take recording off the table entirely and the real question is what an AI tool can do for a motion built around scrubbed-in, hands-busy time. The honest answer is the same shape as any field-heavy motion: the bookends.
Before the case: what did the surgeon flag last time about instrument tray setup? What did the value analysis committee ask for in writing? What's the open item with materials management that's been sitting for three weeks? A rep walking into a case remembering all of it asks sharper questions than one rediscovering the account cold.
After the case: the debrief happens in the parking garage or the car, still in scrubs sometimes, while the details of what the surgeon said about instrument feel or OR staff said about turnover time are still sharp. That window is short — the next case, or the next facility, is already pulling attention — and once it closes, most of what happened in that OR is gone for good.
We've made this case in general terms for in-person sales meetings: no software attends the room, so the tool's job is prep and debrief. Medical device sales is the sharpest version of that argument, because the room isn't just unrecorded — it's often a room the rep's own phone can't enter.
The buying committee doesn't run on a shared calendar
A capital equipment or implant decision in a hospital moves through people who rarely sit in the same meeting. The surgeon's opinion forms during and after cases. The OR nurse manager's concerns come up in scheduling conversations. Materials management asks about cost and logistics in a hallway between two other vendors. Value analysis meets on its own cycle, often monthly, and wants a packet, not a pitch.
Tracking that across facilities — a rep covering a territory might carry six or eight active accounts, each with its own version of this committee — is a memory problem before it's a relationship problem. What the OR nurse manager said she needed for the March value analysis meeting has to survive from the hallway conversation where she said it to the packet that gets built weeks later.
That's a rep-first problem, not a CRM problem: the detail needs to be captured by the person who heard it, in the moment they heard it, and handed back to them exactly when they need it — not routed anywhere else first.
Where Sara fits
Sara is a 24/7 AI deal assistant for reps whose selling doesn't happen where sales software expects it to. She's not in the OR — nothing should be — but she's available the moment you're back in the hallway or the car: call or text her after a case, walk through what the surgeon said about the tray setup or what the nurse manager flagged about scheduling, and it's held for you before the next facility, the next case, the next value analysis packet.
Built for reps carrying a territory of hospital accounts through long, multi-stakeholder cycles. Early access runs through Founders Club, invite-reviewed — apply for early access.
The sterile field sets the rules. The bookends are still yours.
Nobody's asking software to get into the OR, and nothing worth trusting would try. The actual gap is what happens on either side of it — the prep that makes a case go well and the debrief that makes sure what happened in it isn't lost by the next facility on the schedule.
That's the part of the job that's been unsupported for too long. It doesn't need a recorder. It needs somewhere to put what you already know the moment you're free to say it.
By the team at Keel. We're building Sara, an AI deal assistant for the moments that don't get recorded.