From 90% to 99% to 99.9% reliability.
Robots will keep getting better and hardware will keep changing. What will not change is that you need a known quantity of parts each week.
A robot arrives knowing the job in general. It does not know your line.
We build the part that knows your line: your parts, your bins, your light, last Tuesday’s fault. Every mistake gets taught back in. When better robots come along, what we have learned on your line comes with you.
How it works
It earns its way
At the start somebody watches everything it does. The more it proves it can handle, the more it does on its own.
Humans on escalation
When the robot is unsure, it escalates instead of guessing, and a trained operator resolves it remotely, with minimal interruption.
Intervention rates fall
The number we are judged on is how often a human had to step in. Our goal is to bring this number down every week. That curve is the product.
Nothing breaks twice
Every intervention is logged, diagnosed and fed back, so the same edge case does not stop you again. Your parts, your lighting, your tolerances — and a record of every fix.
This is not
a new idea.
This is the system we built and scaled at Onfido, in an industry that made exactly the same journey, from human review to audited automation.
It got there in a particular order: humans did all of it, then the easy cases were automated, then humans kept the hard ones. The accuracy number stayed our problem the whole way. Customers bought the result, not the software.
Robots are at the start of the same curve. Until it plays out, a robot that stops is somebody’s problem. We would rather it were ours than yours.
Got a cell that keeps stopping?
Talk to us