Ask questions about your machines. In plain English.
Use AI-assisted review to ask why a run failed, what would improve performance, or whether a machine is ready for a new task — and get a clear, structured answer.
A few things to know right now
- It works from reviewed records, not uncontrolled live command paths. New runs must be included in the reviewed record before they can be analyzed.
- One machine at a time. Each connection covers a single robot cell, not your whole fleet. Fleet-wide querying is on the roadmap.
- Access is scoped through Xolver rather than published as an open setup recipe.
What you can do with this
Xolver can support AI-assisted review of machine records so teams can understand what happened and what to improve — without touching any machine directly.
The assistant is a review aid, not a controller. It can explain, analyze, and suggest. A human still approves every change, and nothing the assistant says can reach hardware directly.
What you can ask
"Why did this motion stop?"
Your assistant analyses the run record and explains exactly what went wrong — whether it was the AI misreading the scene, a safety rule triggered, or a timing issue.
"What should I change to improve accuracy?"
Based on the failure pattern, your assistant proposes specific changes to how the model is trained — without modifying anything itself. You review the suggestion and decide whether to apply it.
"What tasks should the AI practise more?"
Your assistant suggests which training scenarios to add or change based on where the current model is struggling.
"How should this workcell be laid out?"
Given the current scene geometry, your assistant proposes an optimised layout. It doesn't change anything — you get a suggestion to review.
"Is this configuration ready to deploy?"
Once you've approved a set of changes, your assistant prepares the final input for the training pipeline — ready for the next deployment cycle.
Your AI assistant suggests. You decide.
Everything the assistant returns is a suggestion — never an action. It can analyze, explain, and recommend. But nothing it says reaches your machines directly. A human reviews every suggestion before anything changes.
The AI assistant cannot touch your machines — by design, not by convention.
This isn't a setting you can accidentally turn off. The system is built so that suggestions and actions are completely separate paths — there is no shortcut from a suggestion to the machine.
Access
AI-assisted review is scoped through a Xolver engagement. We review the deployment record, access boundary, and approval path before enabling it.
Implementation details are shared directly with qualified partners rather than published as open marketing copy.
This is a thinking partner, not a robot controller
Your AI assistant can look at what your machines have done and help you understand it. It cannot tell a machine what to do. That distinction is fundamental — and it's enforced in the system, not just as a guideline.
Your AI assistant can
- Explain why a run failed
- Suggest what to change to improve performance
- Recommend which scenarios to train on more
- Tell you whether a configuration looks ready
- Help you understand what happened in a specific run
Your AI assistant cannot
- Send any command to a machine
- Change your safety rules
- Apply any of its own suggestions
- Approve changes on your behalf
- Move any robot, ever
Want to try it with your deployment?
Get in touch and we'll scope AI-assisted review for your deployment records, approval path, and operating boundary.
Get in touchRelated