# From Michelin pass to product launch — the year we built ROA

> A founder note from Yago and Arnav on what the kitchen actually taught us about software.

Mar 19, 2026 · Story · By FoodWeb · 5 min read

Canonical: https://foodweb.ai/blog/michelin-to-product

---

A year ago we were standing in different kitchens, looking at the same problem. Yago was running services at a Michelin-recommended kitchen in Madrid. Arnav was building tools for restaurants who couldn't afford to think about software.

We kept noticing the same thing: every kitchen runs on tribal knowledge. The head chef holds it in their head. The sous chef inherits a fraction of it. When someone leaves, the kitchen forgets.

## What we got wrong at first

We started by building a recipe database. Nobody cared. Chefs don't want another place to type recipes — they want their existing ones to do more work.

So we threw that out and started again from a different premise: *the chef shouldn't have to type anything*. The system should learn from what the kitchen already does.

## What the kitchen taught us

Three things, in order of importance:

1. **Latency is everything.** A tool that's right but slow is dead. We measure UI response in milliseconds because we have to.
2. **Density matters.** Marketing-tool design — big buttons, lots of whitespace — is poison in a kitchen. We pack information.
3. **The agent should be quiet.** ROA only speaks when asked, or when something's about to go wrong. The kitchen has enough noise.

## Where we are now

ROA is in private beta in restaurants across Spain, France, and the UK. The team is two engineers and a chef. We're hiring slowly — only people who've worked in a kitchen.

If you've felt the same friction we've felt, talk to us.