Voice-first hotel operations
As TeamJet’s Founding Product Designer and sole designer, I turned an early-stage idea into a production hotel-operations platform across web, iOS and Android.
The pivotal moment came when my first keyboard-based mobile workflow failed with housekeeping staff. Field research revealed that walkie-talkies worked because employees could use them without removing their gloves or stopping work.
I replaced typed task creation with a voice-to-task interaction that used machine learning to identify the location, task type and responsible department from a spoken request.
We delivered the MVP in three months, signed the first paid SaaS contract after eight months and reached profitability during the first year.
Hotels were paying for complex software while work still happened over radios.
Hotel operations were split between expensive legacy software, paper processes, spreadsheets, phone calls and walkie-talkies. Existing ERP products offered extensive functionality, but hotels used only around 40% of it while still paying for the complete system. The cost extended beyond licences. Dense interfaces required substantial training, slowed employee onboarding and created resistance among frontline workers who needed to complete tasks quickly while moving around the hotel. I joined TeamJet when the product was still an idea and was responsible for defining what it should become. Rather than replacing hotels’ existing property-management systems, we positioned TeamJet alongside them as a focused operational layer. The platform needed to serve very different users: managers monitoring performance from desktop dashboards, supervisors coordinating departments, engineers responding to maintenance issues and housekeepers completing tasks from mobile devices. I owned the product architecture, interaction model and MVP priorities with the founders, engineering and sales teams.
My first mobile interface failed because I designed for a screen, not the hotel floor.
My initial mobile concept used a conventional keyboard-based interface. I simplified the forms, reduced the number of fields and made the controls larger, expecting that a cleaner UI would make task creation practical for frontline staff. It failed during testing. Housekeepers could not comfortably stop work, remove their gloves, hold the phone and type a structured request. The interface was understandable, but it did not fit the physical environment in which it had to be used. It was a difficult result to accept because the design itself appeared logical. My first instinct was to keep refining the interface—larger controls, fewer options and a simpler keyboard flow. Instead, I returned to the hotels and spoke directly with housekeeping and maintenance staff. I learned that the main advantage of their walkie-talkies was not familiarity or speed alone. Employees could press one button, speak and continue working without removing their gloves. The problem was therefore not which UI elements to use. The interaction model itself was wrong.
I stopped refining the UI and changed the interaction model.
I proposed replacing typed task creation with a voice-to-task workflow. An employee could press one large button and say, “Bring towels to room 207,” while the system identified the location, recognised the task type and routed it to the responsible department. Machine learning handled the structured input that the interface had previously demanded from the employee. Instead of asking a housekeeper to select a room, category, assignee and priority, the product extracted that information from natural speech. To test the concept, we purchased inexpensive smartphones with large resistive displays that could be operated while wearing household gloves. The mobile interface was reduced to the essential actions required to record, receive and update tasks. We tested the revised workflow with hotel managers, housekeepers and engineers. Frontline staff strongly preferred the voice interaction because it preserved the simplicity of their existing radios while connecting their work to the wider operational system. This decision changed TeamJet from a digital task-management interface into a product that frontline hotel teams could realistically adopt.
The product reached paid SaaS in eight months and profitability within a year.
We delivered the first MVP within three months, covering the core task-management workflow across the web platform and mobile applications. Six months after the project began, hotels operating under major international brands were piloting the product. The voice-to-task interaction gave TeamJet a distinctive capability and helped demonstrate that digitising hotel operations did not have to make frontline work slower or more complicated. We signed the first paid SaaS contract after eight months and reached the original goal of generating profit within the first year. The work established TeamJet’s product architecture, interaction patterns and design foundations, taking the company from an unvalidated idea to a commercially operating enterprise platform.
