Data & retrieval
Knowledge Base
A structured collection of documents an AI system can search and quote — the source-of-truth corpus that grounds RAG and many AI agents.
01 ——
In plain English
A knowledge base, in AI tooling, is the indexed body of documents an LLM is allowed to draw from when answering. It's the "ground truth" half of retrieval-augmented generation: the model is told to base its answer on what's in the knowledge base, not its training data.
What goes in a knowledge base:
- Internal docs (Notion, Confluence, Google Drive, SharePoint)
- Product documentation and help articles
- Support tickets and resolved cases
- Policy and procedure manuals
- Public web pages a company controls
How it gets indexed:
- Documents are chunked into passages (typically 200–800 words)
- Each chunk is embedded into a vector database
- At query time, the system retrieves the top-K most relevant chunks
- Chunks are passed to the LLM as context for the answer
Why it matters: Knowledge bases are how organisations safely use AI on private data — the model never trains on the docs, it just reads them at query time. Tools like Glean, Notion AI, Microsoft Copilot, and most enterprise chatbots are knowledge-base-driven.
02 ——
Related terms
RAG
Retrieval-Augmented Generation — a technique that gives an AI model access to external documents before it answers, so it can cite real, up-to-date sources.
Vector Database
A database optimised for storing and searching embeddings (numerical representations of text or images) by similarity.
Embeddings
A way of converting text (or images) into lists of numbers so an AI can measure how similar two pieces of content are.
Semantic Search
Search that finds results by meaning rather than exact keyword matches — so "car" finds results about "automobile" too.
AI Search
A search experience that returns a synthesised AI answer with citations — instead of a ranked list of links — powered by retrieval plus an LLM.
Knowledge Graph
A structured representation of entities and the relationships between them — used to give AI systems explicit, queryable facts.
Back to glossaryLast reviewed May 2026