AI that earns trust before it acts
Orbis moves up a ladder only as fast as you allow:
- Shadow: Orbis runs silently and you see its results against actuals
- Suggest: it surfaces suggestions and you approve each one
- Assist: it acts on stable patterns and escalates edge cases to you
- Autonomous: opt-in, for low-risk ingredients only
It learns your kitchen, not an average one
Every forecast correction, prep override and supplier choice is a training event. The model running in your kitchen after a year and a half reflects your guests, your menu and your suppliers, which is something no one starting fresh can copy.
Later, enough kitchens sharing anonymised patterns will let a new restaurant start with a working forecast on day one.
Questions
Will AI replace my chef or manager?
+
No. Orbis prepares decisions and shows its reasoning; people make them. It removes the arithmetic, not the judgement.
What does Orbis need to start?
+
Your POS sales history and your recipes. Orbis reads your current POS read-only and backfills the history; its own till, CASA, is planned for January 2027.
Is my data shared with other restaurants?
+
Your kitchen's model is your own. Network features on the later roadmap use anonymised patterns across kitchens.
Which POS does it work with?
+
Not on day one. Orbis starts by reading the sales history of the POS you already use, read-only, and runs in shadow mode beside it. Its own till, CASA, which runs on any tablet and keeps working offline, is planned for January 2027, so you can switch when you are ready. When you apply, tell us what you run today: Restolution, Trivec / Caspeco, Lightspeed, Toast, Square or another.
How do I try it?
+
Orbis is in active R&D with Helsinki design partners. The MVP is testable from December/January, and R&D access is free during testing.