Mobility AI

Context-driven AI that earns the interruption — intelligence that knows when to wait.

Context

Embedding AI into the in-vehicle experience meant assistance had to surface naturally, without breaking flow or taking control from the driver. The hard part isn't capability — it's restraint: an always-helpful assistant in a moving car is a hazard.

My role

Design lead for the Mobility AI interaction model — defining how AI surfaces, escalates, and hands off, and the rules for timing, confidence, and intervention. Partnered with engineering, product, and data teams on the signals behind those decisions.

AI Surface Engine Live
Mobility AI · Context-Driven Delivery
SPD
AI Decision Log
AI Surface Engine
The AI agent analyzes context and delivers components to surfaces in real time.
Activity Feed

Approach

Two interaction models: AI delivers content through dynamic cards and persistent widgets in high-confidence moments, and hands off to system or third-party apps for deeper workflows. Intervention is gated by "golden moments" — brief windows, read from environmental, behavioral, and intent signals, where assistance is likely useful, welcome, and safe — with criticality levels setting how prominently the AI intervenes while preserving control and reversibility.

The hard call

Restraint as a feature. The system is designed to stay silent far more often than it speaks, deferring or suppressing anything that isn't a genuine golden moment. It's less visibly "smart," but it's what keeps the assistant trustworthy in a safety-critical space.

Outcome

A clear, reusable interaction model for embedded AI: timely assistance without distraction, reliable handoff between AI and applications, and trust preserved through predictability and restraint. This model is the direct predecessor to the Behavioral Orchestration framework.