CartSense: Trust-First AI Reordering
CartSense is a concept AI feature for a grocery delivery app (like Zepto or Blinkit) that helps repeat buyers skip rebuilding the same cart every week. The primary design challenge: build automation people trust, not automation people fear. Full-auto carts risk hidden charges and unwanted items, stripping user control the moment they can't see what they're paying for.
Project intro
CartSense helps repeat buyers skip rebuilding the same cart from scratch every week. The primary design challenge: build automation people trust, not automation people fear. Full-auto carts risk hidden charges and unwanted items, breaking user trust the moment automation touches their money.
My Role
Solo UX/UI designer. I was responsible for research synthesis, flow mapping, wireframing, hi-fi visual design, and prototyping in Figma over two weeks. This is a self-directed concept project—I owned every decision end-to-end.
What I 'have change
Decision 1: Review and approve, not full automation
The Problem: Fully automated background checkouts remove user agency. Users fear surprise charges and unreviewed items.
The Solution: I made the review-and-approve screen the default entry point instead of a background auto-checkout. Every item stays editable with a single tap (remove or swap instantly), and an opt-out switch sits one tap away in settings. Automation removes the manual search-and-reselect work, while control stays entirely with the user.
Decision 2: Visual separation between AI picks and user picks
The Problem: Ambiguity destroys trust. When users can't distinguish between items added by the algorithm versus items they picked themselves, suspicion starts.
The Solution: I gave AI-suggested items a distinct background tag and icon so they are identifiable at a glance and can be reviewed in seconds.
The Problem: A generic "AI added this" label reads as an opaque black box and triggers suspicion.
The Solution: Each suggested item carries a specific line explaining why it's there based on order history (e.g., "Based on your Tuesday morning routine"), building pattern trust over repeat use instead of triggering suspicion on first contact.
Outcome
A dashboard experience that gave users clarity into their spending, helped them act before overspending instead of after, and laid a reusable foundation for future financial features.
What I 'have learned