An inspection report can look precise and still be wrong for the job. If a part is checked against an obsolete drawing revision, the instrument may record the dimension faithfully, the software may produce a polished report, and the wrong conclusion may travel through the organization with impressive speed. That is the failure mode Gleb Tsipursky asks manufacturers to address as inspection becomes more connected and more dependent on AI assistance.
Writing in Metrology News, Tsipursky argues that connected metrology makes this problem more consequential, not less. Digital data moves farther and faster, but speed does not guarantee that a measurement is being compared with the right requirement. His proposed remedy is a "version passport" for every consequential measurement: a compact set of fields that travels with the result and makes its applicable conditions visible to the next decision-maker.
The failure mode is version drift
The article points to persistent identification as part of the technical foundation. Quality Information Framework data can preserve the identity of a measured characteristic across product definitions, inspection plans, and results. The technical promise is traceability. The management challenge is ensuring that each decision follows the correct version of the requirement. Those are related but not identical goals: a system can trace where a number came from without making clear which drawing revision governed the check.
This distinction matters most when AI enters the workflow. An AI assistant can summarize an inspection report, point to an apparent deviation, or recommend which cases need review. But if the tool receives results without version context, it may compare an observation with the wrong specification. The arithmetic can remain correct while the conclusion is invalid. Tsipursky's point is not that AI inspection is unreliable by default; it is that a capable assistant can amplify a bad reference if the reference is not governed.
A passport is a lightweight data contract
The proposal does not require a new software product. It can begin as a set of fields that travel with a measurement:
- the approved drawing or model revision;
- the characteristic identifier;
- the inspection-plan revision;
- the instrument or method used;
- the result timestamp; and
- the identity of the person or system that approved the disposition.
These fields do not replace metrology software, quality systems, or AI tools. They establish a minimum context package so that a downstream reader can see which requirement was in force and who accepted the result. In that sense, the passport is less a new database than a data contract between systems and teams.
Tsipursky links the need to a broader industrial challenge. He notes that the National Institute of Standards and Technology's 2026 smart-manufacturing roadmap highlights integration of heterogeneous data and trustworthy operation as continuing challenges for industrial AI. A version passport targets a specific part of that problem: it makes the origin and applicable conditions of a measurement visible to the next decision-maker.
Test the assistant's judgment, not just its labels
Quality teams can test the idea before integrating a new platform. The article recommends selecting a small set of completed inspection records and deliberately including three cases: one legitimate out-of-tolerance result, one conforming part, and one example with mismatched revisions. An AI assistant is then asked to summarize the evidence, identify the applicable revision, and flag missing provenance. Its responses are compared with an experienced quality engineer's determination.
The evaluation should not stop at whether the assistant labels a part good or bad. Tsipursky recommends recording whether it catches a revision mismatch, whether it invents an unstated tolerance, and whether it can point to the approved source behind a conclusion. Those are the errors that digital handoffs often conceal. A confident answer that misses provenance is the kind of error the test is designed to expose.
Ownership must be settled before the alert
Records will disagree, and the response cannot be improvised during a shipment dispute. The article assigns clear ownership: a quality engineer may determine whether a measurement is valid, design engineering confirms the governing definition, and manufacturing determines whether production can continue. The escalation route should be settled before the alert appears.
This governance detail is easy to overlook because the technical fix sounds modest. A few fields attached to a result may seem like metadata housekeeping. But the fields only work if someone owns the definition, someone owns the measurement, and someone owns the production decision. Otherwise the passport documents a disagreement without resolving it.
The reference needs scrutiny too
Tsipursky's argument differs from the familiar demand for more AI-versus-human comparisons. Even perfect agreement between human and automated inspectors can reinforce the wrong decision if both were given an obsolete specification. The reference itself needs scrutiny. That is a useful warning for quality organizations that may be tempted to measure AI performance only by agreement rates.
Automation can make measurement data travel farther and faster. A reliable chain of custody ensures that the drawing, feature, method, and result still belong together when the decision is made. The next generation of connected inspection, Tsipursky concludes, should make that connection impossible to ignore.