GenAI application for the shopfloor

As an innovation incubator, we pick up on disruptive technologies at an early stage and match them with value-adding use cases. We have used the potential of large language models and complementary methods to develop an AI chatbot in close collaboration with Siemens, Accenture and Volkswagen to support store floor employees in the efficient handling of technical documentation and operating manuals.

Live in action: AI chatbot in production

The advantages of the AI chatbot

Expert knowledge at the touch of a button

Faster troubleshooting

More efficient communication

Upskilling of shopfloor employees

The path to the company’s own AI chatbot

The challenge

The technical implementation of an AI chatbot requires the development of an AI architecture in which LLMs, internal data and AI tools must be efficiently integrated and coordinated. In view of the increasing risk of cyberattacks on AI systems, comprehensive protection is essential. In addition, the underlying knowledge base must be usable in compliance with data protection regulations, while potential risks from AI-generated misinformation must be identified, minimized and actively managed.

The implementation

In order to comply with data protection regulations and the necessary security precautions, the requirements of the EU AI Act must first be met. The appropriate LLM had to be identified for the technical AI infrastructure. An agent-based RAG approach was chosen to merge the internal data and the dialog-based capabilities of the LLMS. In order to reduce incorrect answers as much as possible, it is necessary to build a well-curated knowledge base enriched with metadata. The RAG approach is an efficient alternative to LLM fine tuning to identify relevant data for correct answers, with retrieval augmented fine tuning opening up new possibilities for domain-specific LLM applications.

The result

Using the existing technologies and methodological measures, an AI chatbot was developed within a few months that can provide valid expert information in a two-way exchange with users. The “best-of-breed” AI chatbot architecture can be flexibly scaled and enriched with additional information by connecting external databases. By independently accessing databases, agent-based AI enables interactive, cross-system knowledge processing. Although AI agents can only be used in production environments in a very controlled manner, they offer considerable added value when it comes to expanding the possible applications of AI and overcoming the limitations of RAG.

How do AI agent architectures work?

What our customers say

“Using a GenAI ChatBot in maintenance increases efficiency, reduces downtime and improves decision making. It provides instant access to troubleshooting guidance, predictive maintenance insights and real-time support so technicians can resolve issues faster. By automating routine queries and documentation, a GenAI ChatBot also frees up valuable human resources for a more streamlined and cost-efficient maintenance process. “

João Batata, Technical Manager Body Business Area (AGFKT) | Volkswagen Auto Europa

Quick Win: AI-Chatbots in production