An AI knowledge system in the pilot phase
An ongoing pilot, no final results. This insight shows the approach and the current interim status. We will report success figures only once the pilot is complete.
- Status
- Pilot, ongoing
- Technology
- Azure OpenAI
- Field
- AI enablement
Problem
Two starting points at once. Internal knowledge is scattered across several systems, such as documents, manuals and internal wikis, and staff spend a long time searching for answers. At the same time, a noticeable share of recurring requests, including from customers, could be answered from exactly this existing knowledge but is currently handled manually.
Approach
A RAG system (retrieval-augmented generation) based on Azure OpenAI, connected to a custom retrieval pipeline over the existing knowledge sources. GDPR compliance was part of the architecture decision from the start: data processing stays within the EU, and content is not passed on to third-party models in an uncontrolled way.
Current status
Pilot operation with real but selected knowledge sources. The focus is on answer quality and traceability, meaning which source was used for a given answer. Scaling is not yet a topic.
Outlook
Once the pilot phase is complete, a full case study with solid results will follow. This insight will then be updated or replaced.
The next sensible step
Knowledge scattered across your company?
Do you want to make internal knowledge findable and answer recurring requests automatically?