A digital inspection result can contain an accurate measurement and still be wrong for the job. That happens when a part is checked against an obsolete drawing revision: the instrument records the dimension faithfully, the software produces a polished report, and the wrong conclusion travels with impressive speed. Gleb Tsipursky, a behavioral scientist, argues that AI-assisted inspection makes this failure mode more consequential and that manufacturers should give every consequential measurement a version passport.
The passport is not necessarily 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, the result timestamp, and the identity of the person or system that approved the disposition. Those fields matter because a measurement is only meaningful in relation to a requirement. Persistent identification can preserve the identity of a measured characteristic across product definitions, inspection plans, and results. The technical promise is traceability; the management challenge is making each decision follow the correct version of the requirement.
The list is modest, but each field answers a different question. The revision identifies the governing requirement. The characteristic identifier locates the feature being judged. The inspection-plan revision shows how the check was defined. The instrument or method indicates how the result was produced. The timestamp and approval identity establish when and by whom the disposition was accepted. Without that bundle, a later reviewer may see a number but not the conditions that made the number relevant.
Why AI makes version context more important
AI can summarize an inspection report, point to an apparent deviation, or recommend which cases need review. But a tool that receives results without version context may compare an observation with the wrong specification. The arithmetic can be correct while the conclusion is invalid. This is not a hypothetical edge case in the author’s framing; it is a structural risk in digital handoffs. 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 addresses a specific part of that problem: it makes the origin and applicable conditions of a measurement visible to the next decision-maker.
The point is not that automation is unreliable in general. It is that automation can carry a result farther and faster than the context attached to it. A report may look complete while lacking the revision that governed the inspection. An AI assistant may then produce a confident summary that is internally consistent but tied to the wrong requirement. The missing context is often invisible precisely because the output is polished.
When the version is absent, a reviewer cannot easily tell whether a deviation is real, whether a tolerance was taken from the right source, or whether a conforming result was judged against an outdated definition. The risk grows as results move between systems and teams. Each handoff can preserve the number while losing the requirement that gave the number meaning.
A small test before a platform decision
Quality teams can test the idea before integrating a new platform. The author recommends selecting a small set of completed inspection records. The set should include 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 how often the assistant labels a part good or bad. That metric can hide the errors that matter. Reviewers should record whether the assistant catches a revision mismatch, whether it invents an unstated tolerance, and whether it can point to the approved source behind a conclusion. These are the failures that digital handoffs often conceal. A model that agrees with a human on a simple pass/fail label may still be relying on an obsolete specification, and a human may be doing the same.
The test is deliberately small because it is aimed at provenance, not at benchmark performance. A few records can reveal whether the assistant notices missing version context, whether it treats an unsupported tolerance as established, and whether it can trace a conclusion back to an approved source. Those questions are more useful in this setting than a broad score on good-versus-bad classification.
Ownership when records disagree
Records will disagree, and the escalation route must be settled before an alert appears rather than negotiated during a shipment dispute. A quality engineer may determine whether a measurement is valid. Design engineering confirms the governing definition. Manufacturing determines whether production can continue. Assigning clear ownership when records disagree keeps the version question from becoming an improvised argument under delivery pressure.
This division does not remove accountability; it locates it. The quality function judges the measurement, design engineering confirms the requirement, and manufacturing decides on production continuity. When those responsibilities are clear in advance, a revision mismatch becomes a defined exception rather than a dispute that must be resolved while a shipment is waiting.
This 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. A version passport makes that reference explicit: which drawing, which characteristic, which inspection plan, which method, and which approval were in force when the result was produced.
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 should make that connection impossible to ignore.