A dimensional check can be numerically faithful and still send a shop floor in the wrong direction. If a part is measured against a drawing revision that has already been superseded, the instrument reports the feature correctly, the software assembles a polished report, and the wrong conclusion reaches everyone quickly. In a Metrology News opinion column, behavioural scientist Gleb Tsipursky argues that manufacturers adopting AI-assisted inspection should attach what he calls a version passport to every consequential measurement.
Tsipursky, CEO of Disaster Avoidance Experts and author of The Psychology of AI Adoption at Work (Georgetown University Press, 2026), points to a 26 October Metrology News account of persistent identification, which describes how Quality Information Framework data can carry the identity of a measured characteristic across product definitions, inspection plans and results. The technical promise there is traceability; the management task is making sure each decision follows the right version of the requirement.
The fields that travel with a result
The passport, he stresses, need not be a new software product. It can begin as a set of fields that move with a measurement:
- the approved drawing or model revision;
- the characteristic identifier;
- the inspection-plan revision;
- the instrument or method used;
- the result timestamp;
- the person or system that approved the disposition.
Where AI helps and where it misfires
An AI assistant can summarise an inspection report, point to an apparent deviation or rank cases for review. Given results without version context, however, it may compare an observation with the wrong specification. The arithmetic stays correct while the conclusion becomes invalid. Tsipursky links this to the National Institute of Standards and Technology's 2026 smart-manufacturing roadmap, which lists integration of heterogeneous data and trustworthy operation among the continuing challenges for industrial AI. A version passport addresses one slice of that problem by making a measurement's origin and applicable conditions visible to whoever decides next.
A small trial before a platform commitment
Quality teams can test the idea without integrating anything new. The column suggests taking a handful of completed inspection records, including one genuine out-of-tolerance result, one conforming part and one case with mismatched revisions, then asking an AI assistant to summarise the evidence, name the applicable revision and flag missing provenance. Its answers are compared with an experienced quality engineer's determination.
What matters is not how often the assistant labels a part good or bad, but whether it catches a revision mismatch, whether it invents a tolerance nobody stated, and whether it can point to the approved source behind a conclusion. Those are the errors that digital handoffs tend to conceal.
Who owns a disagreement
When records conflict, ownership has to be explicit: a quality engineer may judge whether a measurement is valid, design engineering confirms the governing definition, and manufacturing decides whether production can continue. The escalation route should be settled before an alert appears rather than negotiated during a shipment dispute.
Beyond human-versus-AI scorekeeping
The argument departs from the familiar demand for more comparisons between AI and human inspectors. Even perfect agreement between the two can reinforce a wrong decision if both were handed an obsolete specification, so the reference itself needs scrutiny. Automation moves measurement data further and faster, and a reliable chain of custody is what keeps drawing, feature, method and result together at the moment of decision, a connection the author says the next generation of connected inspection should make impossible to ignore.