A robot that moves through a room full of people fails in more than one direction. We write each direction as its own line, price it separately, and let you decline any of them except the first.
Injury to people and damage to property that is not yours.
Your unit clips a guest’s foot on a turn and she goes down on tile. Ambulance, X-rays, and the claim her lawyer files four months later.
The robot itself. Replacement, not depreciated book value.
A patron shoves your unit for a laugh and it goes down hard. The optical array shatters. That array is a third of the machine.
Breach response, notification, and privacy liability.
Someone intercepts the optical and audio feed. Your unit has been recording every table in the room, and now so has a stranger.
Harm caused by what the robot said, not what it did.
The onboard model is asked about allergens and answers confidently. It is wrong. Nobody touched anybody, and you are still liable.
Revenue lost while the unit is in for repair.
The unit is fine, eventually. It is also in a crate in Ohio for three weeks and your Friday covers are down forty percent.
The most we pay for any single incident. Bystander is written per occurrence, because two guests going down on two different nights are two separate problems and should not share a ceiling.
The most we pay across the whole policy year, however many incidents. Cyber and AI lines are written in aggregate, because one breach tends to be one event with a long tail rather than many separate ones.
What you pay before we do. Higher deductibles buy lower premiums, and the Downtime deductible is measured in hours rather than money: the first 48 hours offline are yours.
Chassis is the one line whose limit you set yourself. It tracks the replacement cost you declare, and it is priced at one percent of that figure, so a more expensive robot costs proportionally more to cover and the maths stays visible.
Four questions and you have the whole schedule, with every line costed separately.