PATE - AI Bill Splitting App

Designed in Figma in 2024, then built and shipped as a live product. A bill splitter where nobody types a number.

Company

Own product

My Role

Product Designer, UX Research, AI-Assisted Build Direction

Product Designer, UX Research, AI-Assisted Build Direction

Industry

Restaurants & Payments

Date

March 2024 - August 2026

The idea

Remove the maths from the end of a group meal

Remove the maths from the end of a group meal

Remove the maths from the end of a group meal

Build a bill splitter where nobody types a number, and take it all the way to a live product people pay for.

Build a bill splitter where nobody types a number, and take it all the way to a live product people pay for.

Build a bill splitter where nobody types a number, and take it all the way to a live product people pay for.

The story

Designed first, built a year later

Designed first, built a year later

Designed first, built a year later

Every portfolio piece has a tidy origin story. This one is honest: I designed PATE fully before I had any way to build it, and then spent over a year finding out how much a Figma file leaves out.

What the design phase could not tell me was which decisions would actually matter. On the screen, splitting a bill is a clean interaction. In use, it turns out to be a question of trust, of what happens when the input is wrong, and of who at the table can be bothered. Those became the real design problems, and none of them were visible until something ran.

The sections below trace that gap, from the idea to a product people pay for.

PATE: the moment the bill arrives
PATE project timeline
PATE: zero numbers typed

The build

Built alone, directing AI tools

Built alone, directing AI tools

Built alone, directing AI tools

I built the whole thing myself, directing AI tools rather than writing code. That matters less as a technical fact than as a design one: it meant I never handed the work off, so every compromise and edge case stayed my problem to solve rather than someone else's to interpret.

The decisions that follow are the ones I am most willing to defend. Each came from watching the product meet reality, a receipt it could not read cleanly, a person who would not install anything, a flaw I found by attacking my own design. They are small individually. Together they are the difference between something that demos well and something that holds up.

This is the part of the process that no prototype reaches, and the part I think says the most about how I work.

PATE receipt verification
PATE item assignment
PATE claims flow
PATE outcome

What I learned

The interface was the easy part

The interface was the easy part

The interface was the easy part

The thing I did not expect was how much of shipping is judgement rather than craft. Deciding a bill that cannot be cleared is broken regardless of what the code thinks. Deciding what someone's allowance should do when they cancel. Deciding not to launch on a day the numbers told me would go badly, and trusting the instruments I had built to tell me so.

None of that appeared in the original design, and all of it shaped the product more than the screens did. That is the real argument of this piece. The interface was the easy part. Knowing which problems were worth solving, long after the file was closed, was the work.

I wrote up the full build process, the stack and the lessons in how this app was built with Lovable and Claude.

Get in touch today to discuss how I can help you unlock your products 2.0! Let’s get on a call and make it happen with intuitive and impactful design solutions.

Get in touch today to discuss how I can help you unlock your products 2.0! Let’s get on a call and make it happen with intuitive and impactful design solutions.

Get in touch today to discuss how I can help you unlock your products 2.0! Let’s get on a call and make it happen with intuitive and impactful design solutions.