When I joined Omnicell I was brought in to lead SaaS UX. A few months later I was reorged and ended up owning the entire Point of Care product line — including the next-generation Automated Dispensing Cabinet redesign already in motion.

Four teams were working on one device with no shared view of the full experience. Industrial design was an outside firm. Electrical engineering was the bridge between hardware and software. Two scrum teams split pharmacy and nursing workflows. By the time I was in the room, hardware decisions had already been made with UI implications nobody had mapped.

The tracks were being laid while the train was moving. Multiple form factor concepts in play simultaneously — different screen sizes, orientations, surface materials, input methods. None of it decided. All of it consequential for the software.

Redesigning hardware with a decade-long lifecycle isn't a desk job. I went on-site, gowned up, and spent time watching how people actually used this hardware in live hospital pharmacies and nursing stations.

What you see here tells the story before I say a word: a tethered scanner, a keyboard next to the screen, someone reaching up into a cabinet with both hands occupied. Everything observed in those visits shaped every interface decision that followed.

Environmental conditions weren't edge cases — they were the spec. Overhead fluorescent lighting. Noise. Physical display positioning. Gloved hands. Users who are never fully stopped when they interact with the screen.

I built a service blueprint that nobody asked for. Three swim lanes — user actions, hardware, and software — mapped across every workflow the device was part of.

The device had one job: meds in, meds out. A drawer opens, someone puts something in or takes something out, and the system tracks it. Simple mental model. But when you laid the blueprint out, no single team owned that full loop. The hardware team owned the drawer. The software team owned the screen. Nobody owned what happened in between.

That's the seam. The blueprint made it visible for the first time. Every team was suddenly looking at the same device doing the same job — not their piece of it, all of it. Once you can see the seam, you can own it. That's what the blueprint was for. Not documentation. Alignment.

The blueprint told us what questions to ask. The environment told us what the answers had to be. We tested screen sizes and orientations mounted on actual cabinets — not in a lab, on the hardware itself.

We looked at glare from overhead fluorescent lighting. We tested viewing angles because the person reading this screen is rarely standing directly in front of it. We did smudge tests. We looked at touch target sizes for people who weren't always able to stop moving to interact with the screen.

The environment had opinions and we had to listen to them. Our UI didn't always perform. That's what drove everything that came next.

The constraint was an existing design system — Greenlight — built for web and optimized for someone sitting at a desk with a mouse. The response wasn't to abandon it. It was to extend it.

Same tokens. Same components. But a kiosk-specific layer built on top. The environment is a design input, not a constraint to solve after the fact — and the design system had to encode that understanding by default, so any screen built on top of it inherited those constraints automatically.

Responsive web design wasn't the answer here. Responsive web design assumes a browser. This was a kiosk bolted to a wall in a medication room, operated by someone in gloves who has somewhere else to be in thirty seconds.