Increased efficiency in Product Lifecycle Management through AI

Complex systems, high effort for users

PDM and PLM systems, as well as enterprise tools in general, have evolved over decades. Functionality, UI concepts and process logic have been continuously expanded without making the overall system simpler. Everyday PLM work is characterised by complex topics such as change and release processes. Many tasks are repetitive, rule-based and information-intensive, while others require the analysis of complex structures with variants or multi-level bills of materials. This results in very high demands on employees.

AI provides relief across the entire value chain and offers clear advantages in engineering as well. Engineers typically have strong domain-specific expertise but are not PLM specialists. Their focus is on design and engineering work, not on spending unnecessary time navigating historically grown system logic.

Significantly simplified user interaction

An AI agent enables PLM functions to be accessed via natural language. Users can describe their needs in a “human” dialogue, while the agent executes the underlying processes deterministically. This ensures system integrity while significantly simplifying interaction for users.

In a concrete customer context, Intelliact developed a demonstrator to illustrate the potential of AI in Product Lifecycle Management. On the one hand, the goal was to identify critical topics and challenges involved in building such an AI application. On the other hand, the aim was to generate ideas for various PLM use cases. The underlying AI framework or LLM was specifically conditioned to act exclusively as an assistant for PLM-related tasks.

Valuable support for PLM-specific tasks

In Windchill, a widely used PDM/PLM system that manages product data, bills of materials, documents, changes and release processes and serves as a central engineering data source in many industrial companies, the agent provides valuable support. Typical tasks handled by the AI agent include searching for or creating parts, retrieving multi-level BOMs, creating and modifying BOMs, and supporting check-in/check-out processes. Mandatory attributes, contexts and default values are automatically taken into account.

Deterministic processes and probabilistic interpretation clearly separated

For use in PLM environments, a clear separation between deterministic and probabilistic AI is essential. Tasks that change system states—such as creating, checking out or releasing objects—must remain deterministic, reproducible and auditable. Interpretative tasks, such as understanding incomplete input, identifying relationships or structuring unstructured information, can be supported probabilistically.

An AI agent can therefore interpret user intent, but must execute system-required actions with technical precision. For effective support of PLM processes, the targeted combination of both approaches is decisive.

AI in product structures and migrations

AI provides additional value during migrations or when cleaning up large product structures. Direct inference based on existing, correct data enables a proprietary, finely tuned model that provides context-specific support, for example in identifying alternative parts or cleaning up inconsistent BOM variants.

Alternatively, cloud-based LLMs can be used to deliver quick results, though they are less specialised and impose higher requirements regarding data protection and cost control. Regardless of the approach, the quality of training and contextual data remains a decisive factor.

Data Cleansing mit KI effizienter gestalten

Bei Maschinen oder Anlagen mit vielen Varianten können hunderte tausend BOM-Relationen entstehen und nicht alle lassen sich rein regelbasiert prüfen. KI-Modelle unterstützen dabei, Muster zu erkennen, Beziehungen zu validieren und unklare Strukturen korrekt einzuordnen. Dies verkürzt Analysezeiten und reduziert Fehlerquellen, ohne deterministische Regeln zu ersetzen.

Conclusion

For AI to provide targeted support in PLM and deliver sustainable efficiency gains, a deep understanding of architecture, processes and product structures is required. AI expertise alone is not sufficient, nor is PLM expertise in isolation. Only the optimal combination of both enables the development of specific solutions that operate reliably, remain auditable and noticeably relieve users.

Intelliact brings these competencies together and develops AI applications that deliver practical, secure and user-oriented value.

Further application areas in the PLM context

AI is particularly well suited for repetitive, knowledge-based standard tasks and offers considerable potential in various areas, including:

• Efficiency improvement: automated change notice triggers, classification
• Decision support: release readiness, alternative parts
• Quality and compliance: automated flagging, optimisation
• Knowledge management: supplier integration, information loops
• Predictive intelligence: performance tracking, risk management