Today, building a grocery cart on Instamart means searching for each item, picking the right variant, and tapping "add" — over and over. This case study reimagines that flow with AI: users describe their whole order in plain language — by text or voice — and an AI assembles the cart for them.
The two proposed solutions
Solution A — Natural Language Cart Creation. Users type all their items in one go, e.g. "1 kg potatoes, 2 one litre milk packets, Colgate toothpaste." The AI parses items and quantities, applies sensible defaults where a quantity is missing, flags out-of-stock items (and can suggest alternatives), and presents a clean, editable list to confirm before ordering. (Impact: High · Effort: Medium)
Solution B — Conversational & Voice-Based Ordering. Users speak or chat their order in a back-and-forth flow. The AI asks follow-up questions for missing quantities or brands, confirms the full order at the end, and shows a visual list for final review and edits. (Impact: High · Effort: High)
The deck
Prototype
We built a working prototype — Instamart Smart Cart — on Lovable to demonstrate both the Type Order and Voice Order experiences:
Try the Instamart Smart Cart prototype →
A couple of notes on the prototype: for Type Order, separate items with a comma. For Voice Order, the demo hardcodes the resulting order to illustrate the expected behaviour — in a real build, a speech-to-text model would transcribe the user's speech, and an LLM would turn that unstructured transcript into a structured item list.