A measurement can be numerically correct and still produce the wrong decision. If a part is checked against an obsolete drawing revision, the instrument may record the dimension faithfully, the software may generate a polished report, and the team may reach an invalid conclusion quickly. The problem is not arithmetic. It is the absence of version context around the result.
That risk becomes more consequential as metrology systems become connected. Gleb Tsipursky, a behavioral scientist and author, argues that manufacturers adopting AI-assisted inspection should attach a version passport to every consequential measurement. The proposal is less a new software platform than a discipline: ensure that each result travels with the identity of the requirement it was checked against.
Traceability is technical; version control is managerial
Persistent identification through 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 same characteristic can be followed through different stages of the quality workflow. The management challenge is different. A traceable record is not automatically a correct decision. The decision must follow the correct version of the requirement.
Connected metrology makes the failure mode more important because results move farther and faster. A measurement taken against one revision can be summarized, forwarded, and acted on before anyone notices that the governing definition has changed. Digital handoffs can hide the mismatch rather than surface it.
The fields that make a measurement self-explanatory
A version passport does not need to be a new 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 prove that a part is good. They make the origin and applicable conditions of a measurement visible to the next decision-maker. That visibility is the point. A quality engineer, an AI assistant, or a production manager should be able to see which requirement the result was evaluated against and whether the provenance is complete.
Why AI assistance raises the stakes
AI can summarize an inspection report, point to an apparent deviation, or recommend which cases need review. Those are useful functions. 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. The assistant may also fill gaps with an unstated tolerance or present a confident recommendation without exposing the approved source behind it.
The National Institute of Standards and Technology's 2026 smart-manufacturing roadmap identifies integration of heterogeneous data and trustworthy operation as continuing challenges for industrial AI. A version passport addresses a specific part of that problem. It does not solve data integration or model trust in general. It makes the applicable conditions of a measurement explicit enough for the next decision to be checked.
A small test before a platform decision
Quality teams can test the idea before integrating a new platform. The test does not require a large dataset. It requires records that expose the relevant failure modes. A team could select a small set of completed inspection records and include at least one legitimate out-of-tolerance result, one conforming part, and one example with mismatched revisions.
The team can then ask an AI assistant to summarize the evidence. The instructions should require the assistant to identify the applicable revision and flag missing provenance. The responses should be compared with an experienced quality engineer's determination. The comparison should not be limited to how often the assistant labels a part good or bad. That metric can miss the most important errors.
Instead, the team 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 errors that digital handoffs often conceal. A system that agrees with a human on a pass-fail label may still be relying on the wrong specification. Agreement is not evidence of a valid reference.
Ownership must be settled before the alert
When records disagree, the organization needs clear ownership. A quality engineer may determine whether a measurement is valid. Design engineering may confirm the governing definition. Manufacturing may determine whether production can continue. These roles are related but not interchangeable, and the escalation route should be settled before an alert appears rather than negotiated during a shipment dispute.
The version passport supports that division of responsibility by making the relevant evidence visible. It does not remove the need for judgment. It gives each function a shared basis for deciding which question it is answering: Is the measurement valid? Is the requirement current? Can production proceed?
Beyond the AI-versus-human comparison
The argument differs from the familiar demand for more comparisons between AI and human inspectors. 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 confident human and a confident model can share the same blind spot when the governing drawing or plan is out of date.
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. The version passport is a practical way to start: not a claim that every measurement is correct, but a requirement that every consequential measurement be explicit about the version it was measured against.