Contacts
Get in touch
Close

Contacts

Romania, Cluj-Napoca – 400157
Str. Ploiești Nr. 9

Romania, Vama – 447350
Str. Fermei Nr. 14

+40 774 463 429

info@inteliscience.net

PMI Authoring Automation for 300+ Automotive Designers

Cases

PMI Authoring Automation at a European Automotive OEM – MBD Rules, NX Open Automation and a Method 300+ Designers Work To

PMI authoring automation replaced the manual part of model-based definition for more than 300 design engineers at a European automotive OEM. inteliScience SRL built the MBD rule set, the NX Open automation behind it and the PMI methodology the design organisation now works to — over 18 months, across two Siemens NX version changes.

§ 01 — Abstract SECTOR Automotive USERS 300+ designers DURATION 18 months

More than 300 design engineers at a European automotive OEM author PMI to an automated rule set built by inteliScience SRL, and the client design methods organisation reports that PMI authoring previously took about 40 percent longer than it does today. The work covered three things that only pay off together: the MBD rules themselves, NX Open automation that applies them inside Siemens NX, and a written PMI methodology the organisation can maintain. The client describes the result as lower cost, more consistent models, and a concrete step toward full MBD.

LAST UPDATED AUGUST 18, 2026 · FIGURES RELEASED BY THE CLIENT FOR ANONYMOUS PUBLICATION

§ 02 — Record
Client
European automotive OEMNOT NAMED
Sector
Automotive
Users
More than 300 design engineers
Scope
MBD rules, NX Open automation, PMI and MBD methodology
Duration
18 months
Environment
Siemens NX, two NX version changes during the project
Our role
Rules, automation and methodology — inteliScience SRL

The situation: model-based definition moved the work, it did not remove it

The design organisation was moving toward model-based definition. That move deletes the drawing, but it does not delete the annotation work — it moves it into the model. Instead of dimensioning views, designers author PMI: dimensions, tolerances, datums, surface finishes and notes, each of which has to satisfy the works standard and the internal methodology before a model is released.

Done by hand, that work is both slow and inconsistent. It is slow because a designer authors the same recurring callouts on every part, one at a time. It is inconsistent because two experienced designers will annotate the same feature slightly differently, and both versions look correct on screen. For industry context: in a 2022 survey by parts data vendor CADENAS covering more than 100,000 engineers and designers, nearly half reported spending at least an hour every working day searching for parts alone — routine lookup is a large, well-documented share of a design engineer’s day, and PMI authoring is exactly that kind of routine.

Inconsistency is the expensive half. Downstream, PMI is not decoration: CAM, CMM programming and inspection read it. A callout that is not semantically bound to its feature is invisible to all of them, and the error surfaces late — in manufacturing, not at release.

What changed in the working day

§ 03 — Before+40% effort
  1. manualAuthor every recurring callout by hand, part after part
  2. manualLook up the applicable works standard rule while authoring
  3. by eyeCheck completeness and semantic binding visually before release
  4. variesTwo designers annotate the same feature two defensible ways
  5. lateErrors surface downstream in CAM, CMM or inspection
Afterbaseline
  1. by ruleRecurring callouts are generated from the MBD rule set
  2. encodedThe works standard lives in the rules, not in memory
  3. uniformThe same feature produces the same PMI, on every part
  4. judgementDesigners spend their time on functional tolerancing
  5. documentedMethodology is written down and maintained by the client

What the client reports

The figures below come from the client design methods organisation and were released for anonymous publication. They are what that organisation observes in its own work, not a controlled study — the stamps say so.

§ 04 — Measured delta

Released by the client design methods organisation for anonymous publication.

Effort to author PMI on a model

index 140 index 100 about 29% less

client-confirmed The client reports PMI authoring previously took about 40 percent longer than today, for the same tasks in the same design organisation. Shown as an index because 40 percent more time before corresponds to about 29 percent less time after.

Design engineers working to the automated rule set

0 at project start 300+ rolled out

client-confirmed Rollout across the design organisation, not a pilot group.

Consistency of PMI across parts

designer-dependent rule-defined step toward full MBD

client-confirmed Qualitative statement, no metric attached — described by the client as better quality and a concrete step toward full MBD.

How does automating PMI authoring actually work?

Automating PMI authoring works in three layers, and skipping any one of them is why such projects usually stall. The rules have to be written down before they can be coded, the automation has to live on supported Siemens NX extension points so it survives upgrades, and the methodology has to belong to the client so it outlives the project.

§ 05 — Three layers
Layer 1 · Rules

The works standard becomes an explicit rule set

Before anything can be automated, the conventions have to exist as rules rather than as habits: which callout applies to which feature class, which default tolerance the works standard prescribes, how a note is phrased, what a released model must contain. Writing that down with the client method owners was the first half of the project and the part that made the rest possible.

OUTPUT — an MBD rule set the client owns and can extend
Layer 2 · Automation

NX Open applies the rules inside Siemens NX

The automation recognises the feature, selects the applicable rule and authors the PMI semantically bound to the geometry, in the right view, phrased to the works standard. It runs on the NX Open API — a supported Siemens extension point rather than a parallel island tool, which is why two NX version changes did not end the project.

OUTPUT — recurring callouts generated instead of typed
Layer 3 · Methodology

A written method the organisation maintains itself

Automation without methodology decays: the rules drift, nobody knows why a callout looks the way it does, and the next standard revision breaks it quietly. The project therefore delivered the PMI and MBD methodology as documentation and training for the client method owners, so the rule set is maintained inside the organisation rather than by us.

OUTPUT — 300+ designers working to one documented method

What the solution is built on

§ 06 — Built on

Supported Siemens extension points and the client’s own standards. Nothing that dies with the next NX upgrade.

CAD Siemens NX client environment
Automation NX Open — feature recognition and rule-based PMI authoring source to the client
Rules MBD rule set derived from the works standard and PMI conventions owned by the client
Method Written PMI and MBD methodology, plus method-owner training maintained by the client

Related work on the same ground: MBD and PMI compliance automation for Siemens NX and NX Open custom tool development.

Two NX version changes in 18 months

Over the 18 months of the project the client moved through two Siemens NX version changes. For automation that touches PMI, that is the moment such tools normally die. This one did not, for one unglamorous reason: it is built on the NX Open API rather than on recorded interactions or undocumented internals, so a version change meant a regression run and a handful of adjustments instead of a rewrite. Any team evaluating NX automation should ask about this specifically — a tool that cannot survive an upgrade is a rental, not an asset.

What we would do differently

We would spend longer on the rule set before writing code. The first automation rounds encoded conventions that the method owners had described confidently but that turned out to have exceptions nobody had documented — a feature class where a different tolerance applies, a note phrased differently for one product line. Each exception surfaced as a bug report against the automation when it was really a gap in the rules. Running a short, deliberate rule review with two or three real part families first would have moved that discovery weeks earlier and made the automation phase calmer.

When does automating PMI authoring pay off?

Automating PMI authoring pays off when the same callouts are authored repeatedly against a written standard, by enough people that inconsistency has a cost. In our experience the case is strong when a design organisation runs model-based definition seriously rather than as a pilot, when recurring feature classes make up a large share of annotation work, and when downstream consumers — CAM, CMM, inspection, suppliers — actually read the PMI. It is a poor fit when the works standard exists mostly as tribal knowledge: the honest first project is then writing the rules down, and automation comes after.

MBD consulting and NX PMI training cover the two steps before automation: a written rule set and designers who apply it.

§ 07 — Questions

Frequently asked questions

Why is the client not named?
The client is a European automotive OEM whose works standards, PMI conventions and NX Open code are confidential. Naming the company alongside a description of how its annotation process used to work would be commercially damaging for them, so this case study publishes the approach and the released figures instead. The same anonymous label is used throughout.
Does the automation replace the designer?
No, and it is not meant to. The automation authors the recurring callouts that follow a written rule — standard features, works-standard notes, default tolerances. Functional tolerancing and the datum scheme remain engineering judgement, which is precisely where the freed time goes. The designer releases the model and stays responsible for it.
What happens at the next NX version change?
A regression run and, in this project, a handful of adjustments. The automation is built on the NX Open API, a documented and supported Siemens extension point, so an upgrade is a maintenance task rather than a rebuild. This project went through two NX version changes and kept running.
How long does a project like this take?
This one ran 18 months, from the first rule workshops to a rule set used by more than 300 designers. The length was driven by the rule work and the rollout, not by the coding: writing down conventions that had lived in people, agreeing them across product lines, and training method owners takes longer than automating them once they are agreed.
When would this not be worth it?
When the rules are not written down and nobody has the mandate to decide them. Automation makes a rule set faster; it cannot invent one. It is also a poor fit for a small design team whose annotation work has little repetition, because the rule and methodology effort then outweighs the authoring time it saves.

Your designers’ judgement is excellent. Stop spending it on repeat callouts.

Send us one part family and the section of your works standard that governs it — we will show you how much of that PMI is rule-governed, on your own models.

Talk to an MBD Automation Engineer