AI Knowledge Base for Machine Manufacturer
Development of an AI-powered knowledge database that captures, structures, and makes the expert knowledge of long-standing employees accessible to all staff via an intelligent chatbot.
AI Knowledge Base for Machine Manufacturer
Artificial Intelligence
The Challenge
For projects in this vein, our AI knowledge base service offering describes how we scope, build and support comparable deliveries from Germany.
A mid-sized machine manufacturer with 180 employees faced a critical problem: Within the next 5 years, 12 key employees with an average of 30+ years of operational experience would retire.
This knowledge – from machine settings to troubleshooting to customer insights – existed only in the heads of these experts. Previous documentation attempts with SharePoint and Word documents had failed due to lack of usage and missing structure.
The challenge: How can implicit expert knowledge be systematically captured and prepared so that new employees can quickly access it?
Our Solution
Solution screenshots
We developed an AI-powered knowledge database based on RAG (Retrieval Augmented Generation). In structured knowledge workshops, we captured the expert knowledge of experienced employees – through interviews, process observation, and documentation of decision paths.
The content is stored in a vector database (pgvector) and made semantically searchable via GPT-4.
Employees can ask natural language questions through a chat interface such as 'How do I set up the CNC mill for aluminum?' or 'What do I do when customer XY complains?'.
The system finds relevant knowledge blocks and generates context-aware answers with source references. A feedback system enables continuous improvement of answer quality.
Results
After 6 months of use, measurable results are evident: The onboarding time for new employees was halved from 6 to 3 months. Recurring support requests to experts decreased by 65%. The system now contains over 2,400 knowledge blocks from 8 departments.
User satisfaction is at 4.6 out of 5 stars. Particularly valuable: Even after the first 3 experts left, their knowledge remains fully preserved and accessible.
Features
Feature overview
- Natural language search via chat interface
- RAG technology for context-aware answers
- Source references and links to original documents
- Structured knowledge capture through workshop methodology
- Automatic categorization of new content
- Feedback system for quality improvement
- Role-based access rights (GDPR-compliant)
- Integration into existing intranet
- Offline-capable desktop application for workshop floor
- Multilingual (DE/EN) for international locations
- Automatic updates on process changes
- Anonymized usage statistics for knowledge gap analysis
FAQ
Frequently asked questions about the AI knowledge base for mechanical engineering
When does a RAG knowledge base pay off for mechanical engineering companies?
How does the system handle multiple manual versions and PDF sources?
Can a mechanical engineering knowledge base reduce hallucinations?
Which teams benefit most in mechanical engineering?
Transparency about this case study
So the statements above can be judged properly, we disclose what kind of project this is, what the results are based on and who reviewed the text. More on our project approach and an overview of all reference projects.
- Case type
- Client project, anonymised or shown under a project name – Real project; company name, industry details or individual figures are generalised at the customer's request.
- Measurement basis
- Comparison of onboarding time for new staff and the number of recurring expert enquiries before and after the knowledge base went live.
- Measurement period
- Six months of production use after roll-out
- Data source
- Figures reported by the department heads plus anonymised usage statistics from the system (knowledge blocks, user ratings).
- Scope of the figures
- Figures are rounded and stripped of identifying details; the order of magnitude is preserved.
- Publication status
- Evidence pending in the new approval register
- Approval scope
- The existing anonymised publication remains available; case-specific approval evidence still needs to be recorded in the new register.
- Evidence record
- Internal project file and anonymised reference record.
- Technical review
- Björn Groenewold, Managing Director of Groenewold IT Solutions GmbH and Hyperspace GmbH –
Change history
- Evidence details added to “AI knowledge base in mechanical engineering”: case type, measurement basis, data source and technical review.
- Results and solution description of “AI knowledge base in mechanical engineering” revised; the German version was aligned.
- Case study “AI knowledge base in mechanical engineering” published.
Project Details
Client
Northern German Machine ManufacturerCompleted
2024
Technologies
Client Testimonial
"We used to answer the same questions 10-15 times daily. Now colleagues ask the chatbot first – and it usually knows the answer better than I do because it also has the knowledge of my colleagues. The project has taken away our fear of knowledge loss."
How Sandra finally stopped asking the same questions every day
Sandra has been in production planning for 8 months.
She used to spend at least an hour every day looking for colleagues who could help her – often they were in the workshop or in a meeting. 'I constantly felt like I was bothering everyone,' she recalls.
Since the knowledge database has been running, Sandra simply types her question into the chat. Most of the time she has the answer in under 30 seconds – including info on who originally contributed the knowledge. 'The best part is.
I can look things up at 6 in the morning when nobody else is around yet.
And I learn in the process because I see why something is done a certain way – not just how.' Her onboarding time was significantly shorter than for colleagues who started before her.
Team Voices
"Finally, I no longer have to explain 15 times a day how the special setting for the French plant works. The system explains it exactly as I would have – just more patiently."
Michael K.
Machine Operator, 28 years of experience
"When I heard that my knowledge was going 'into a computer,' I was skeptical. But the workshops were really good – they understood what we do, not just superficially. Now I'm proud when I see that my tips help others."
Petra S.
Quality Inspector, 25 years in the company
"The collaboration with Groenewold was different from other IT companies. They didn't just program but were genuinely interested in our work. You can tell from the result."
Jürgen H.
Workshop Manager
Partnership Instead of Project
- 1We took the time to truly understand the people and their work – not just check off requirements.
- 2Regular workshops with the experts where we came as listeners, not as know-it-alls.
- 3Even after go-live, we stay on board: Monthly check-ins, continuous improvement of answer quality, expansion to new departments.
- 4The client has a dedicated contact person who knows the system and the team – no anonymous ticket system.
More References
Planning a similar project?
Use our interactive cost calculators for an initial estimate – free and non-binding. Or schedule a consultation directly with our experts.