Zorveta One

AI for restaurants

AI safety: how AI is kept in bounds

The limits on AI in Zorveta One: confirm before placing, role permissions, human approval for money and an audit log.

AI in Zorveta One stays inside limits: it asks the guest to confirm, it follows role permissions, money-moving actions need a human and every action is logged. Rules are enforced by the order engine, not by the assistant, so a clever message cannot talk its way around them.

When you would use it

  • “A guest types a message that tries to change the rules of the assistant.”

  • “A staff agent suggests a refund and you want to see who approved it.”

  • “An owner asks what happens if the AI is wrong.”

How it goes, step by step

  1. 1

    Same engine

    AI uses the order engine, not a side door.

  2. 2

    Same permissions

    An assistant has the rights of its role and no more.

  3. 3

    Human on money

    Refunds and price changes go to a manager.

Settings and edge cases

AI safety settings and edge cases
TopicWhat to know
Audit logEach AI action is logged.
Turn it offSwitch any AI channel off per location.
Reporting a problemContact details for reporting are on the security page.

Questions about this feature

Can the AI change prices or give discounts?

No. Those go to a manager.

Can I read the eight questions to ask?

Yes, in the article on AI ordering safety.

See it on your own menu.

Bring your menu. We will show how Zorveta One fits your restaurant.

Book a demo