Scope of Work
Where This Work Sits
This is one of several initiatives I’ve led across two tenures at Genesis Healthcare: Product Designer from 2019 to 2022, Senior Product Designer from 2024 to now. The design system I built during the first tenure is what the platform described here was built on.
Specific product screens are protected under NDA. What follows is an approved summary, with representative figures standing in where real interfaces can’t be shown.
The Problem
Every System Had Its Own Front Door
Genesis Healthcare operates 290+ facilities across the country, and when I started, every department ran on its own system, with its own login and its own password-reset ritual. Clinical staff were mentally exhausted before they even reached a patient.
Training deadlines went out by email and routinely died in junk folders, which is also roughly how benefits worked: they went unused because employees never learned they existed. Password resets alone were enough to swamp IT support. I was the only designer on the project, working in 4–6 week release cycles.
The Research
Clinical Staff Spent More Time Navigating Systems Than Using Them
I ran click-tracking across the existing tools, interviewed clinicians and directors, deployed facility-wide surveys, and benchmarked enterprise healthcare platforms. Being a licensed OT practitioner was a real research advantage: I already understood these workflows from the inside, so I could tell the difference between what staff said slowed them down and what actually did.
Every method landed in the same place. Staff already had the tools they needed. Those tools were scattered across systems with no shared logic for where anything lived, and people burned whole stretches of their day just hunting. The thing to fix was navigation.
The Other Problem
Training Compliance Was Falling Through the Cracks
Required courses were their own mess. Between Continuing Education Credits (CEUs) for state licensure renewal and mandatory company trainings, there was no way to see how many CEUs you had or which trainings you’d actually finished. Even telling required courses apart from optional ones took guesswork, and the deadlines lived in those same junk-folder emails.
Staff wanted their licenses current, and the company badly wanted its trainings completed. What was missing was any single place where either side could see the state of things.
The Pivot
The Platform That Shipped Began as an Off-Hours Prototype
My research pointed to one unified platform. Leadership wasn’t sold on going that far, so the first funded direction was a lighter “hub”: a landing page that linked out to all the existing systems without replacing any of them. I designed and piloted it.
It tested poorly. Staff called it “not an improvement”, and they were right: it added a layer on top of the mess. So I quit trying to win the argument in a deck. I built a working prototype of the unified platform on my own time and demoed it to stakeholders directly.
Seeing it worked better than describing it. The unified approach became the funded direction, and it’s what shipped.
The Platform
Everything Behind One Login
Single sign-on shipped first, alone, months before anything else. One authenticated entry point replaced dozens of separate logins. That sequencing was deliberate: clinical staff have no spare attention for a tool that rearranges their day mid-shift, and one botched rollout would have confirmed every suspicion that this was just another system to endure. SSO touched all 44,000 people while changing nothing about how they worked, which made it the cheapest goodwill we could buy.
The new information architecture came straight out of the click-tracking data. Tools are grouped by clinical workflow, so the categories match the questions staff walk in with on an ordinary day. The reworked IA and the education module piloted in a small set of facilities, where we watched real usage and fixed what confused people before going wide. Legacy links stayed live the whole time; nobody hit a hard cutover.
Formal training for 44,000 people was never going to happen. Each facility instead had a couple of designated go-to staff with early access and a direct line back to us. Questions stayed local and answers stayed fast, which is how things spread through a facility anyway: someone leans over and shows a coworker.
The AI Assistant
The Assistant Had to Earn Its Place
The centerpiece is a custom AI assistant I designed and helped architect, and it had to clear an awkward bar first: my own research said staff needed fewer things to learn, and an assistant is one more thing. It paid its way by behaving like navigation. It knows where you are and what your role is, and it surfaces what you were about to go hunting for. Most days the answer is there before you’ve finished working out what to ask.
It answers policy questions, walks staff through benefits enrollment, explains CEU requirements by state, and routes complex issues to the right department. Every response cites its source, so staff can see exactly where an answer came from before they act on it.
The design question I spent the most time on was what happens when it’s wrong. In a clinical organization a confident wrong answer costs more than no answer, so below a set confidence level the assistant refuses to guess: it says so, and routes the question to the right person with context attached. It also launched deliberately narrow: policy, benefits, and CEU questions, nothing clinical. Scope widened release by release as the accuracy record allowed. HIPAA compliance, role permissions, and activity logging were requirements from the start; the harder design problem was teaching the thing to know its own limits.
The EdTech Solution
CEU Tracking That Finally Works
I designed a dedicated education module with one priority: get people to the right course in as few clicks as the platform could manage. Research showed staff wanted courses sorted by highest CEU value first, so they could knock out annual licensure requirements quickly. That became the default sort.
For staff in states and disciplines that require an ethics CEU, the assistant identifies the requirement automatically and pins the relevant course to the top of the list.
Progress tracking sits front and center, with completion status and upcoming deadlines visible without leaving the page. Deadline nudges from the assistant close the loop that email never could.
Results & Impact
reduction in IT support tickets
improvement in training compliance
decrease in system navigation time
Approximate, directional figures shared with permission; exact numbers are confidential. Each one has its own source. The ticket number is help-desk volume year over year, and it moves with single sign-on. The compliance number is course completion inside the education module. Navigation time comes from re-running the original click-tracking study on the new IA. I distrust platform-wide numbers that can’t say which change did the work, so these were kept separable on purpose.
The response that mattered most was qualitative. Staff said the platform finally felt built around their actual day, and in a clinical setting that kind of trust is what decides whether a tool gets used at all.
Reflection
The Test I Still Use
Every AI feature had to pass one test before it shipped: would a knowledgeable coworker do this better? If yes, it wasn’t ready. That test is the reason the assistant leans on context over clever prompting. A good coworker already knows your role and your deadlines, and doesn’t wait for the perfect question.
Being both the designer and a former clinician cut both ways. I didn’t have to translate terminology or spend weeks earning rapport, but I also knew the workflows well enough to assume pain points the data didn’t support. Click-tracking is what kept me honest.
Next time I’d instrument measurement from day one. The results were strong, but cleaner per-feature attribution would have told me precisely which change drove which gain. An A/B framework would have sharpened every decision after it; the organization wasn’t ready for one, and I should have pushed on that earlier.