Plant Simulation is based on modular building blocks that can be flexibly extended using the proprietary scripting language SimTalk and—more recently—also Python. However, the system’s high performance comes with a corresponding level of complexity. 

In practice, simulation models are therefore often developed by specialized service providers. Due to the division of labor between plant operators, OEM specialists, and external engineering service providers, consistent standardization of modeling components is essential. 

The Automotive Library establishes a binding industry standard in this regard. It provides preconfigured modeling components for key types of equipment in the automotive industry, including conveyor systems, machining production lines, body shop cells, and paint shop systems. 

The goal of the copilot integration was to make this structured knowledge available more quickly, contextually, and in a user-oriented manner.

The Automotive Library includes heterogeneous data sources such as training materials, guidelines, source code, and documentation in various formats (PDF, PowerPoint, HTML). To make this content usable for a large language model, systematic data modeling was required. 

The focus was on introducing a consistent hierarchical heading structure and semantically segmenting the content. The objective was to explicitly represent logical relationships and enable context-aware querying by the copilot.

Today, the copilot supports users in getting up to speed with Plant Simulation more quickly. It facilitates re-onboarding after project breaks and accelerates operational modeling—especially when working with SimTalk. 

At the same time, the next stage of development is becoming clear: while the copilot effectively supports users in operation and functional understanding, the conceptual structuring of complex simulation models still remains with the user. The cognitive translation of source data into a consistent model continues to be a domain-specific expert task. 

Looking ahead, software-agnostic copilots are conceivable—systems that do not primarily support the use of a specific application, but instead draw on domain-specific expertise. Such systems would need a deep understanding of production systems, simulation methodologies, and experimental design in order to go beyond pure functional assistance and deliver real substantive value.