How do demonstrators become product functions, and how do they feed into the products and development processes that companies work with today? That is the question this last part of the whole series answers. To this end, it brings the three chapters together. What is important to emphasize is that the demonstrators do not become isolated AI features, but additional functions precisely where PROSTEP is strong anyway: in traceability across system boundaries and in the integration of complex PLM landscapes. The first two chapters mark out the lines along which these capabilities will feed into the products.

From Traceability to OpenCLM

The traceability assistant presented in chapter 3, part 2 is only a preliminary stage. It finds and substantiates connections wherever data is properly maintained, or where its content can be made accessible. The next step goes beyond that: with APL, missing references between data that today lie in separate silos can be systematically reconstructed, marked, and made available as checkable proposals. For this we use a combined search across the various stores in which these data sit: classical databases, graph databases, full-text and vector indexes.

Only this makes the approach of data linking, as it is conceived in OpenCLM, implementable: relationships across system boundaries are no longer presupposed as though they had long been cleanly maintained but are opened with AI support. A vision of end-to-end traceability thereby becomes a realistic path, provided that the data sources are connected, the search spaces prepared, and the proposals made verifiable.

Agent-Supported OpenPDM Connectors

The second line concerns the integration of PLM systems itself. PLM integrations rarely fail because a single field cannot be read. The difficulty lies in the domain mapping: which objects belong together? Which structure must be transferred from the source system into the target system? And which rules apply in between?

This is precisely where the planned extension of the OpenPDM connectors with AI agents comes in: complex mapping processes such as the mapping between systems can thereby be developed step by step with AI. The technical basis is the BIBO interface (Basic In Basic Out, a generic import and export format on which PROSTEP’s PLM connectors are based). Based on this interface, an agent uses or generates domain integration operations, for example traversing an object tree or assembling a structured information element from many individual objects. Such building blocks come about at runtime when needed, or come from a library of proven operations.

The appeal lies in the fact that the agent works at the level of domain operations, instead of struggling with many machine-level individual interfaces that would overwhelm it with deeply structured data models. For this to be viable, such scripts are not uncontrolled one-off actions: they have to be visible, checkable, and versionable, and released depending on the case of use. That is exactly the idea of controllability that runs through the whole series.

The Same Methodology for Products and Services

The products and services of PROSTEP are based on AIND: on the one hand, we develop our products ourselves with AIND. The methodology described in chapter 1 drives the further development of OpenCLM, OpenPDM, and our connectors. On the other hand, we deploy AIND to realize our customers’ PLM integrations: the same methodological discipline and the same agent-supported toolkit with which we build are available to our customers as a service.

The Circle of the Series Closes

Here the circle of the whole series closes. The same methodology (AIND as described in chapter 1) and the same language and the same agent (APL as described in chapter 2) become applications (chapter 3) with product functions that have their strengths where PROSTEP is strong. AI is shifting the limit of what is possible in PLM, because PLM is to a large extent engineering methodology cast in software.

Talk to us about which traceability and integration tasks in your system landscape can benefit first from AI-native product capabilities.