»The collaboration with the Fraunhofer IAO helped us to test our task quickly and in a structured manner using state-of-the-art technology. The IAO was able to guide us on how to document implicit knowledge appropriately and integrate existing information so that it could be used for automated evaluation. They also identified the limitations of the approach and the need for changes to the customer documentation. The team quickly understood our complex support processes, supported the preparation of our existing documents and integrated them into a test system. It was particularly positive that technical feedback from the support team was incorporated directly into the further development of the solution.«
Andreas Wagener
Dr. Fritz Faulhaber GmbH & Co KG
Contact at the AI Innovation Center
Chandan Kumar
AI-Powered Assistance for Industrial Drive Systems
Initial situation
Dr. Fritz Faulhaber GmbH & Co. KG develops high-precision drive systems for demanding and critical applications. The technical support team handles complex customer inquiries on a daily basis, which often require several queries and steps to resolve.
To do this, employees have to search through many distributed sources of information (manuals, short notes, checklists, question trees, mapping tables), each of which only applies to certain products or interfaces (e.g. RS232, CAN, EtherCAT).
This costs time and delays responses. In addition, cases that have already been resolved are not systematically documented.
Solution idea
An AI-supported application helps the support team to systematically clarify the context and create answers.
The AI analyzes the request, summarizes the problem in an understandable way, identifies relevant documents and creates a proposed solution. Employees can adopt or adapt this directly.
Incomplete or incorrect answers make gaps in the documentation visible. Curated solutions are stored in a knowledge database so that similar cases can be solved more quickly in future.
Benefit
By automating research and suggested answers, the support team is relieved of routine tasks and can concentrate more on complex cases.
Response times are shortened, response quality increases and customers receive precise support more quickly. Ideally, an initial response with an inquiry or attempted solution is provided at the speed at which out-of-office responses would otherwise be made.
This improves the availability and trouble-free operation of the drive systems, increases customer satisfaction and strengthens Faulhaber's competitiveness in the long term.
Implementation of the AI application
An assisting ticket management system was developed for a proof of concept. The application structures support requests, extracts, for example, controllers and the motor to be connected, summarizes the context and suggests answers based on the existing documentation. The selection of sources and the quality of the output can be evaluated more efficiently via a user interface. The focus during the progress period was on the technical implementation of the RAG approach (Retrieval Augmented Generation) with a knowledge database. Further work was financed by third-party funds.
The challenges in adding the manuals and other sources to the knowledge base were:
- Cleanup and preparation of documents, e.g. removal of PDF footers.
- Selection of suitable chunking strategies, e.g. parsing as markdown and keeping related sections in the same chunk.
- Enrichment with metadata, e.g. product families, interfaces for pre-filtering relevant sources.
Ideas for improving the search include
- Generate sub-queries for each sub-problem of a customer inquiry.
- Hybrid RAG: Combination of vector search and classical methods (e.g. BM25).
- Agentic RAG techniques: dynamically adapt search strategies to the query type.