ETIHADBEYOND BORDERSFeatures 01 & 02 · Chauffeur & AI Flight Search
An expert usability inspection of both prototype flows — the Business-Class chauffeur booking & live tracking, and the AI-powered flight search — assessed against Nielsen’s 10 usability heuristics with severity-rated findings, the issues resolved through iteration, and a prioritised roadmap.
I inspected each flow screen-by-screen against Nielsen & Molich’s ten usability heuristics, rating issues on the standard 0–4 severity scale. This is an expert inspection by the designer — a fast, structured way to surface problems early. It complements, but does not replace, testing with real users (see the validation plan in the roadmap). Strength scores below are a relative judgement per heuristic, where 5 = exemplary.
No usability concern
Fix if time allows
Low priority to fix
Important to fix
Must fix before release
Both flows are mature and share one design system. The AI search funnel and the chauffeur’s reassurance model are the standout strengths; remaining issues are mostly cosmetic or stem from intentionally stubbed secondary flows (live data, payment, voice).
The chauffeur flow turns a logistics task into a premium, reassuring experience — the right emotional register for Business Class.
Every step earns trust: progress dots, a recommended pickup tied to the flight, “We’re watching your flight,” a driver-assigned card, and Uber-style live tracking with an ETA countdown. The premium promise is felt, not just stated.
A tight-time selection (preset or custom) triggers a clear “that’s cutting it close” warning, a time that would miss check-in is blocked with a plain explanation, and cancelling asks for confirmation; exceeding 3 guests/bags auto-upgrades to a BMW X7; an unavailable action (check-in) is visibly disabled rather than tappable.
Direction-aware copy (“collect you from home” vs “meet you at the airport”), plain-language timing (“arrives airport ~15:00”), and a tangible vehicle card map directly to how travellers think about an airport transfer.
Optional choices (special requests, custom time) are progressively disclosed, keeping each screen focused. The aesthetic reads first-class without clutter.
aria-disabled).The AI flow’s defining quality is that it is explainable and keeps the traveller in control — the AI does the heavy lifting, the user makes the call.
The system shows what it understood (parse chips), why it recommends each trip (“why this trip… $88 under your $700 budget”), and lets the user steer at every turn — refine chips, edit criteria, exact-date calendar, time slots, fare and extras, save, and undo. “AI proposes, you decide” is delivered, not promised.
Results are a ranked shortlist by destination with a clear rationale — a deliberate contrast to OTA aggregator density (coupons, lock-price, dozens of combos). This is the feature’s core value and its strongest usability decision.
An AI “thinking” state, “searching 40+ routes,” match scores, and a total that updates live as the user changes time, fare or extras keep the system’s state continuously visible.
When nothing fits the budget, the screen doesn’t dead-end — it states the lowest achievable price and offers one-tap fixes (raise budget, fewer travellers).
Evaluating the flows together surfaces a clear cross-cutting strength: they read as one product.
Identical tokens, type (Figtree + IBM Plex Mono), components (cards, steppers, sheets, CTAs) and the real Etihad tail across both features — so moving from search to a Business-Class service feels seamless. (Heuristic 4.)
Both flows “recommend first, then let the user control”: a smart default (best-value flight / flight-matched pickup) with progressive disclosure for power options. Users learn the pattern once.
Live-region announcements, focus moved to each screen’s title, role=radio/switch controls, reduced-motion handling and generous tap targets are applied consistently in both prototypes.
Plain-language reassurance over hype; no dark patterns, countdown pressure or fake scarcity — appropriate for a premium airline and consistent across features.
Sequenced from low-effort polish, to validation with real users, to the engineering needed for production.