// CASE 08 · EFM · MYSPOT
An assistant for the boring part of the working day.
eFM digitises buildings and the services inside them. MySpot was its virtual assistant for the small, recurring, low-value tasks people at work do every day: book a desk, a room, a service. We researched, designed the experience and built the design system.
1
primary action per screen, by rule
5
items in the bottom bar, the same on iOS and Android
0
free-text replies in the chatbot: every answer is a guided choice
1
style guide with documented code, handed to the developers
01
Executive summary
After a round of qualitative research with the people who would use it, we designed MySpot around one idea: the app should disappear behind the task. Navigation in a bottom bar, content revealed a little at a time through cards, gestures kept to the ones people already know, and a chatbot that never pretends to be human. Every screen has one primary action, placed where the thumb is. The whole thing shipped as a documented design system so eFM could keep building without us.
02
Four phases
➔ Qualitative user research on recurring tasks in smart working
➔ UX design: navigation, content model, conversation design
➔ UI and interaction principles: one action, one finger, familiar gestures
➔ Design system and style guide with code documentation
Tools & tactics
Qualitative research
Conversation design
Mobile UI patterns
Design system documentation
03
Issues discovered
➔ Recurring low-value tasks were scattered across tools, emails and people.
➔ Finding a thing fast was the whole job, and the navigation did not make it fast.
➔ A chatbot with open answers generated errors and frustration.
➔ Empty results left people in front of a blank screen.
➔ Without documentation, every new screen drifted from the last.
Five principles for a workplace app
One primary action per screen
One-finger use, thumb zone first
Reveal a little at a time
The bot is a bot, and says so
Document it or it does not exist
04
Results
eFM got an app experience, a design system and a style guide detailed enough for developers to build from without asking. What I cannot show: adoption numbers inside the client companies stayed with eFM. What went wrong, in a useful way: the guided-only chatbot limits freedom, and we chose that on purpose, because a bot that answers a little worse but always correctly beats one that answers brilliantly one time in three. What stays: it was my first design system delivered as code documentation rather than a Figma file, three years before Ekitai made that the default.
05
Lessons learned
➔ Familiar patterns beat clever ones when the goal is speed
➔ Constrain the bot and you free the user
➔ An empty state is a design decision, not an absence
➔ A design system is finished when the developers stop asking