Spliit — AI bill splitting in seconds
Group expenses are tedious to reconcile. Spliit needed a flow where you scan a receipt, assign items, and settle up — without the spreadsheet.
An AI receipt-scanning flow with item-level assignment and instant settlement, designed to feel fast and trustworthy from the first scan.
Overview
Splitting a group bill is a small task that everyone hates. Someone ends up squinting at a receipt, typing items into a spreadsheet, and chasing friends for the difference. Spliit wanted to collapse that whole ritual into a few taps: scan the receipt, tap who had what, settle up. Our job was to make that flow feel fast and trustworthy from the very first scan — because a bill-splitting app that gets the math subtly wrong is worse than no app at all.
Where AI earns its keep
This is a textbook case of AI in the loop rather than AI as the product. The scan reads the receipt and proposes the line items; the person stays in control of the result. That framing drove the design. We treated the model's output as a draft, not an answer — every extracted item is editable, totals are always visible, and the interface makes it obvious when something needs a human glance. The goal was never a magic trick. It was to remove the typing while keeping the trust.
The tradeoffs we made
Receipts are messy — crumpled paper, faded thermal ink, currencies and formats that vary. Rather than pretend the OCR would be perfect, we designed for the cases where it isn't. Low-confidence reads surface for confirmation instead of silently entering a wrong number. Correcting an item is one tap, not a restart. That honesty about the edges is what makes the fast path feel safe enough to actually use.
What shipped
Spliit went from concept to a working 0→1 product: the receipt-scanning flow, item-level assignment, instant settlement, and a landing page to put it in front of users. The same team carried it end-to-end, so the AI design decisions and the product decisions never drifted apart.
