Fleet Redeployment Hub

Challenge

An enterprise fleet client needed to validate whether AI-assisted interaction patterns could reduce a multi-hour redeployment workflow into a guided, auditable experience. Existing operations depended on fragmented spreadsheets, email threads, and manual status checks, creating high cognitive load and low confidence in action outcomes. The PoC needed to prove a natural language interface could support high-volume decisions while clearly communicating intent, system state, and action confidence.

My Role

• Solo designer on the PoC: owned all UX, UI, and design system decisions from concept through handoff-ready prototype
• Design-to-development bridge: translated Figma designs into KendoReact implementation using AI-assisted front-end tooling (Lovable) and VS Code
• AI workflow evaluator: assessed AI-generated UI implementations against design intent, establishing ground-truth corrections and quality criteria
• Documentation author: wrote copilot-instructions.md and component specifications enabling the engineering architect to build accurately from design output

Approach

• Designed a four-state natural language command bar (idle, processing, results, error) that communicates AI confidence and action scope clearly
• Built a semantic status badge system with consistent color semantics (blue/gray/orange/red) that communicates vehicle availability at a glance across a dense inventory grid
• Established vehicle grid with filtering, bulk selection, and batch redeployment actions, designed for operators managing hundreds of assets
• Designed a side drawer for individual vehicle detail and a batch redeployment modal for multi-vehicle action confirmation
• Used AI-assisted tooling (Lovable + Figma REST API) to generate and evaluate front-end implementations, directly informing what AI-generated UI gets right and where it needs human correction

Impact

• Delivered a client-ready PoC that validated AI-assisted redeployment workflows in a single operational interface on a compressed timeline
• Defined natural language interaction patterns with explicit idle/processing/results/error states so operators could interpret system status and next actions quickly
• Designed confidence-aware communication patterns that made AI output actionable by clarifying scope, certainty, and recovery paths when errors occurred
• Created reusable KendoReact patterns and implementation guidance that improved engineering handoff quality and reduced interpretation risk
• Demonstrated that a natural language + bulk-action model can replace multi-step coordination loops with a faster, lower-friction decision flow

Tools & Technologies

Figma · KendoReact · Lovable · VS Code · React

Links