WatchLock

WatchLock —
AI-powered cargo theft prevention

A 2013 IoT fleet security product — hardware digital locks on liquid cargo tankers with GPS and geofencing — redesigned for 2026 with an AI intelligence layer that shifts fleet managers from reactive firefighters to proactive supervisors.

ROLE

Lead UX · End-to-end

PLATFORM

iOS + Android

DURATION

8 months

STATUS

Shipped

✦ Predictive halt scoring

✦ CV tamper detection

✦ Explainable AI

✦ Volume forecasting

GPS + geofencing

Multi-drop trips

92%

Parent alert accuracy after AI tuning

3x

Faster incident response vs manual review

0→1

Full product designed and shipped

01 — THE PROBLEM

Parents worry about their teenage drivers but existing apps either overwhelm with raw GPS and data or send alerts too late. The challenge: how do you give families real safety insight without making the app feel like surveillance.

02 — PROCESS

1

Discovery

12 parent + teen interviews. Mapped anxieties and trust gaps.

2

Define

3 personas, journey maps, core use cases prioritised.

3

Design

Lo-fi → hi-fi. 4 concept directions, 2 rounds of usability testing.

4

AI layer

Designed risk score UI, explainability pattern, alert logic.

5

Ship

Handoff, design QA, and post-launch iteration.

03 — AI FEATURES DESIGNED

AI DESIGN DECISIONS

These weren’t backlogged features — each AI capability was designed with a specific user anxiety in mind and paired with an explainability pattern so users could trust the output.

Trip risk scoring

ML model scoring each trip 0–100 on speed, braking, distraction signals. Designed the score card, trend chart, and “why this score” explainability drawer.

Driving behaviour forecast

Predicted weekly behaviour drift based on time-of-day patterns. Designed the weekly digest card and proactive nudge notification system.

Autonomous incident playbook

When risk threshold triggers: auto-notifies parent, suggests a check-in call, logs the incident. Designed the escalation flow and override controls.

Explainable alert UI

“Hard braking detected near school zone at 7:42pm” — not just “bad trip.” Designed the alert anatomy to always show the signal, not just the verdict.

04 — KEY SCREENS

Home / garage

AI risk history

Trip history

Alert explainer

Family members

05 — WHAT I LEARNED

The hardest UX problem wasn’t the AI — it was trust. Parents needed to understand *why* an alert fired before they would act on it. Designing explainability as a first-class surface (not a footnote) was the decision that most improved usability test scores. The AI is only as good as the UI that frames it.

NEXT PROJECT

AI-Triaging

View

Create a free website with Framer, the website builder loved by startups, designers and agencies.