Home / Case Studies / AI Knowledge Base for Equipment Group

AI Knowledge Base for Equipment Group

30k+ docs via RAG, after-sales response up 60%

↑60%

After-sales response efficiency

3万+

Technical documents connected

92%

Weekly active engineers

18→6个月

New-hire ramp-up

Background

The client is a large domestic equipment manufacturing group with more than 800 after-sales engineers nationwide. Three decades of technical manuals, repair cases and parts catalogues were scattered across multiple systems and shared drives, totaling more than 30,000 documents. Field engineers spent an average of 40 minutes finding information, and new-hire onboarding took 18 months.

Challenges

Document formats were extremely complex, mixing scanned PDFs, CAD drawing notes and Excel parameter tables. Core product parameters meant data could not leave the intranet. After-sales work is safety-critical, so every AI answer had to be traceable to a page in the source document.

Solution

An FDE engineer worked on-site for three months. The first two weeks were spent shadowing frontline troubleshooting to map 12 high-frequency question types. Four weeks were used for OCR and layout analysis of scanned documents, chunking content into a three-level product, component and fault structure. An open-source LLM with RAG was deployed on intranet GPUs, and every answer included a source page citation. The system was piloted with 120 engineers in East China, iterated weekly, then rolled out nationwide and integrated with WeCom.

Client feedback

The ZhiMaBox engineer spent three months in our after-sales center and understood frontline usage better than we did. Every AI answer can be traced back to the original page, and our senior technicians trust it.

Have a similar project in mind?

Get a Free Assessment →
💬