Grounded by design: how myExpert AI stays inside the OEM's knowledge
An AI assistant on a factory floor is only useful if it is trustworthy. Here is the constraint we put on myExpert from day one, and why we think it is the right one.
There is a version of an AI maintenance assistant that sounds impressive and is quietly dangerous: one that answers any question by drawing on the whole internet. On a factory floor, that is exactly the wrong design. A confident answer pulled from a generic source can send a technician down the wrong path on the wrong machine.
So we built myExpert AI with a hard constraint. It answers questions only from the documents the OEM has uploaded for that specific machine. No third-party data, no open-web guessing.
Why the constraint is a feature
The instinct is to see this as a limitation. We see it as the entire value. The OEM already has the authoritative knowledge: the manuals, the parts catalogue, the service history, the checklists. The job is not to replace that knowledge with something broader. It is to make that exact knowledge instantly accessible, in any language, at the moment a person is standing in front of the equipment.
This also keeps the OEM in control. They decide what myExpert can see, per installation. The assistant never says more than the OEM has chosen to share, which means the OEM's voice and the OEM's accuracy are what the customer experiences.
What this looks like in practice
A technician scans the QR on a machine, asks a question in their own language, and gets an answer in seconds, drawn from that machine's documents and nothing else. It behaves like the OEM's best service engineer, cloned into every customer's pocket, and it is fully secure because it only searches within the uploaded material.
Grounding is not a guardrail we bolted on later. It is the premise the whole feature is built on.