Overview

experimental UX concept · 2026

What if Beli made deciding where to eat easier?

What if Beli… had a ‘merge’ mode for deciding with friends? (๑ᵔ⤙ᵔ๑)

Also an attempt to adjust my design workflow with AI !

role
UX Designer (personal project)
timeline
April 2026
tools
Figma, Loveable, Chatgpt, Claude

The Problem

Picking a place with friends is weirdly hard even when you both already have options saved.

and then you end up getting Chipotle… wasting the potential of trying a new place in the city.

Research

Most food apps act as if:

“The more options you show, the better the choice.”

But in group decisions:

  • More options = more negotiation, not more clarity.
  • People default to the loudest preference instead of the best one.

Not optimizing for choice … but for agreement.

Why this matters

I started by prompting ChatGPT and Claude to help me articulate why this felt like a real problem worth solving -> and kept coming back to the same tension: it's not a discovery problem, it's an agreement problem.

Flow Challenge

Option 1

**Discovery First** • Let users browse restaurants • Then help them narrow together

vs

Option 2

**Alignment First** • Collect preferences up front • Then surface options that match both

I used Claude to pressure-test both directions, prompting it to argue for Discovery First so I could see the strongest case against my instinct. Alignment First still won.

The Journey

Remember the number of minutes wasted picking a movie for movie night.

I have seen that deciding where to eat is as difficult: different food preferences, dietary restrictions, etc.

Without a full research team, I used Claude to simulate the process: generating personas, mapping pain points, and walking through journey stages.

The biggest pattern it surfaced: the frustration doesn't start at the restaurant. It starts at “so where do you want to go?”

That reframed everything.

AI as Part of My Process

I used ChatGPT and Claude to pressure-test my thinking, prompting them to argue against my decisions before I committed to them.

I also built a clickable prototype in Lovable early on, which made friction in the flow obvious in a way static frames don't.

Using AI earlier meant fewer surprises later. ₊˚ ⋅

Final Designs

The final flow focuses on getting to agreement quickly.

Users start a Let’s Eat session inside Beli, invite a friend, and immediately see overlapping saved spots.

Design system reference: Beli's Design System by Nisheta Gupta, Shivani Mehta, Hugo He and Sandra Ye

Reflections

Hope you enjoyed this quick case study! ₊˚ ⋅ 𓐐𓎩 ‧₊˚ ⋅

I love Beli’s 🍽️₊˚ simplicity and how intentional it is about keeping user journeys short.

Using AI throughout changed my process more than I expected. It replaced the whiteboard conversations I'd usually have with a team… and pushed me to question assumptions earlier.

Designing for alignment instead of discovery changed how I think about group decision flows. (๑ᵔ⤙ᵔ๑)

My beli heheh: beliapp.co/app/souupy

  • A golden savoury crêpe in a paper cone held up in front of the La Crêperie de Hampstead stall, under a round lamp.
  • Two hot pots on tabletop stoves: one in front with sliced meat, cabbage, mushrooms, crab stick and sausage in a clear broth, with udon and rice beside it.
  • A tall parfait in a wine glass: strawberry pieces and matcha ice cream under berry scoops, pink macarons, flowers and red sugar glass.
  • A red table set with charred tandoori chicken on a metal platter with onions and green chutney, plus bowls of curry, chickpeas and rice.

My ultimate favourite foods from my travels in London, Toronto, New York and DC!