AI Assistant for Technical Support
Machine Learning
Large Language Models (LLMs)
Retrieval-Augmented Generation (RAG)
- Our client is a Swiss-based manufacturer of specialized measurement devices.
- They needed an AI solution to efficiently organize the knowledge base and make technical resources easily accessible to technicians in real-time.
- We conducted in-depth interviews to understand workflows, data sources, and user needs, forming the basis for a well-defined project scope and system architecture.
- To help the client decide whether to scale the project, we developed a set of documents considering a comprehensive project overview such as a project roadmap, technical workfow, user path, and a budget projection, along with an extender PoC to validate the technical solution.
- The client was highly satisfied, appreciating the pilot's functionality and needs-driven design.
Our client, a global manufacturer of specialized measurement devices, faced a significant challenge in managing their technical documentation and internal knowledge base.
With a strong commitment to long-term product support, they needed an efficient way to handle and share "tribal knowledge" - information often siloed within teams or individuals.
Their ideal solution was an AI-powered assistant that leverages Retrieval-Augmented Generation (RAG) to organize and deliver technical resources like manuals, service logs, and historical repair data to technicians in real-time.
User needs and approach
To ensure the system addressed real-world needs, we conducted a series of in-depth interviews with management and end-users.
It gave us a clear picture of workflows, the existing technical infrastructure, data sources, and the hierarchy of relevant documents.
We shaped the project scope, focusing on essential requirements to launch a functional pilot.
With this foundation, the client could justify the business case for a knowledge assistant while we developed a detailed system architecture and roadmap outlining manpower requirements and a phased budget.
Technical solution
We opted for RAG over traditional fine-tuning due to the domain-specific nature of the documents. RAG allowed the AI to dynamically access knowledge sources and offered several advantages:
- Easily manage the retrieval database, e.g., incorporate new manuals as new device models are released.
- Improved response accuracy through comprehensive embedding, minimizing errors.
- Using prompt engineering to tailor responses, such as delivering structured answers with direct document citations.
Application Workflow
The application’s workflow consists of three core components:
-
Document Ingestion: Technical documents, such as PDFs, are converted to text, processed, and stored as embeddings.
-
Knowledge Retrieval: The AI identifies the most relevant documents for each user query, synthesizing information into concise, accurate responses.
-
End-User Interaction: The AI assistant provides step-by-step instructions and specific document references, enabling technicians to quickly access critical information.
We delivered an interactive Proof of Concept (PoC) system, which included a UX-tested interface, a scaling plan, a detailed roadmap, and a budget projection. The client expressed high satisfaction with the pilot’s capabilities and appreciated the thorough needs assessment that informed its design.