Every department in this hospital ran bookings differently, and there was no way to stop two people from double-booking the same slot. I turned the capacity rules into an interface staff could read at a glance during a busy shift. It shipped 69 requirements, 29 of them entirely new, and the hospital reported full adoption after launch.
AFTER LAUNCH
Full
staff adoption
Formally reported by the client
- Delivery and adoption
69 requirements delivered, 29 with no equivalent in the old system. The client formally reported full staff adoption following launch. - Account growth
Design direction on this account extended across six projects in total: this flagship system, plus equipment maintenance, patient admission, medication verification, doctor billing, and a healthcare CRM platform. - Business outcome
The trust built through this engagement opened the door to ExaApp's later move into a government-level hospital partnership.
This project replaced a Hong Kong hospital’s internal booking system: one that had never been built around how this hospital’s staff actually worked. The hospital’s trust in this work led to 5 more projects and opened the door to a government-level hospital partnership. I was part of the project from the pitch stage, as the platform grew into an AI-assisted product.
Project Setup
My job was making this system usable under time pressure for three tiers of users sharing one platform: system administrators, department heads, and front-line staff. The rules, from booking capacity per timeslot to when a conflict should warn rather than block, were defined by the product manager with the wider team: the client, the developers working out technical implementation, and me. Working from them, I designed how each rule appeared on screen, so every tier could act without stepping on another’s bookings.
I built the product’s design system, core screens, and interaction framework with later releases in mind, not only the first. When AI-assisted features were added, they were built on that same foundation rather than as a separate layer.
The Staff's Pain Point
The hospital's existing system had been built for no one in particular. Staff worked around missing features by repurposing unrelated ones, and four problems kept surfacing:
A double-booked timeslot only became visible after the fact.
Department heads could not adjust their own service types, resources, or timeslot rules. Every change, however minor, routed through IT.
Departments had no consistent way to control how many bookings could land on the same resource at the same timeslot. Overbooking was either invisible or handled inconsistently, department by department.
The old system simply did not match how this hospital ran its departments day to day.
None of these was the hard part. The real challenge was making the fix work for departments that operated nothing alike on one shared platform.
From Logic to Interface
The system’s logic rests on two capacity thresholds: each resource, at each timeslot, carries a Load Level and a Max Load Level.
Bookings under the Load Level go through with no warning. Bookings between the two trigger an overridable warning. Bookings past the Max Load Level are blocked outright. A separate resource-blocking tool lets a department head take a resource offline for a repair or closure. Front-line staff simply see no open slot, with no need to know why.
That logic sits on a small-core structure: the calendar (create, edit, reschedule an appointment) as the core, with department settings, service types, and message templates layered on top as extensions rather than separate systems bolted on. I designed how each of those extensions surfaced within the interface, so a change in one department’s rules did not ripple into another’s.
WCAG AA was a starting constraint on the design system from day one, not a fix added later.
The design system’s typography, components, and interaction patterns were set for a staff population with a wide age range and long shifts at a screen. Font size, contrast, and touch target sizing were chosen with that range in mind, and visual clutter was kept low on purpose.
Key Decisions I Made
Reversing the calendar's core layout after testing
The calendar, the product's core screen, needed a layout staff would use dozens of times a shift. My first direction was a Multi-Column Timeline / Grid Aggregation View: a denser, more modern way to see multiple resources at once.
I ran an A/B test during a design walkthrough session, comparing that layout against a Columnar Calendar / Day Grid View closer to what staff already knew from the old system. End users did not take to the newer layout: it asked them to relearn a screen they used constantly. I reverted to the columnar, day-grid approach.
The A/B test result was the deciding factor, not a preference call. Familiarity outperformed a layout that looked more sophisticated on paper.
Designing for shared access without shared risk
A midpoint review put 17 people in the room: IT, technical leads, eleven user representatives, and one decision-maker who could send the project back to the drawing board. Two interaction risks surfaced there. Staff covering another department still needed to see their own department's bookings without leaking access to bookings outside their scope, and a single accidental drag on the calendar could silently move a patient's appointment.
For the first, I designed an account structure where one login holds a list of accessible locations; a dropdown lets staff switch which location they are currently viewing, and a booking outside that list shows as a blank, unreadable card until the right location is selected. For the second, I added a confirmation step before any reschedule saves, catching accidental drags before they reach the patient record.
IT, technical leads, and the decision-maker called the system more dependable than the previous vendor. User representatives called it clean and easy to navigate. The review did not end the project. It confirmed the direction, and the work continued.
Designing Across Many Stakeholders
This project had to answer to the client, IT, technical leads, user representatives, and a decision-maker, while three tiers of staff, each working differently, used it every day. The calendar A/B test showed me that among all those voices, a layout that looked more sophisticated was not the one that worked better. The deciding measure was how daily users actually performed.
That distinction is easy to state in principle and easy to miss in the moment, especially when a new direction looks like an obvious improvement. It shaped how I approached the rest of the interface: default to what staff already trust, and save the less conventional choices for the places where the old system had no equivalent to fall back on.
*Additional detail available on request.