Seeing Machines
Guardian Live
Making the world's roads safer by turning six fragmented tools into one platform that fleet managers actually trust
Embedded on-site with Seeing Machines' development team, I was responsible for the end-to-end design direction of Guardian Live, including establishing the visual language, building a 200+ component design system, writing user testing scripts, and driving an agile transformation across a 20+ person development team.
~90%
Customer satisfaction (up from 48%)
2–5 min
Intervention time (down from 5–10 min)
8 billion+
Kilometres tracked
The problem
Seeing Machines' AI-powered devices detect driver fatigue using in-cab cameras and computer vision. The detection technology was world-class. The tools fleet managers used to act on that data were not.
Over the previous decade, six separate systems had been built in isolation. Fleet managers juggled multiple tools to monitor safety, respond to incidents, and generate reports. Their largest client managed 8,000+ trucks across 40+ countries. Intervention management took 5–10 minutes as managers navigated between systems. Onboarding required multiple training sessions. Consolidated reporting was effectively impossible.
The brief: Unify six systems into a single platform that gives fleet managers real-time visibility, fast incident response, and reliable reporting.
Research and discovery
Direct user access was restricted by client agreements, so I designed a research approach around the constraints: six to eight collaborative workshops with fleet managers, regular stakeholder interviews, and approximately 20 rounds of prototype testing (3–5 users per round) conducted by Seeing Machines' internal Human Factors team using scripts I wrote.
Three insights shaped the direction:
- Fleet managers thought in terms of scope, not systems. They wanted data filtered by service provider, account, or fleet. The existing tools forced constant context-switching.
- Incident response was the most critical workflow, and it was too slow. The multi-system process stretched intervention times to 5–10 minutes, a gap that could mean the difference between a near-miss and a serious accident.
- Reports were structured wrong. The team assumed a visual refresh of email reports would improve adoption. User feedback revealed the real problem: reports were organised by event type, while fleet managers thought in terms of vehicles. That insight shifted the redesign from cosmetic to structural.
