Rows of reflective cubes in monotone
← All workshops

Course title

Enterprise Knowledge Retrieval with Generative AI

Grounded AI assistants built on organisational knowledge

Overview

Retrieval-augmented generation enables a language model to answer from an organisation's own documents rather than from its general training. This workshop demonstrates how to build, evaluate, and maintain a knowledge assistant whose answers can be traced to authoritative sources.

A general-purpose chatbot has no knowledge of your policies, courseware, or regulatory circulars, and may fabricate answers with apparent confidence. Retrieval-augmented generation addresses this by locating the relevant passages first and instructing the model to answer from them alone.

The session moves from document preparation, chunking, and indexing to retrieval tuning, grounded generation, and citation, and concludes with the methods used to measure and sustain answer quality over time.

Learning outcomes

By the end of the workshop, you will be able to

  1. 01

    Explain the architecture of retrieval-augmented generation and determine where it is the appropriate solution.

  2. 02

    Prepare, segment, and index organisational documents for accurate retrieval.

  3. 03

    Build an assistant that grounds each answer in cited source material.

  4. 04

    Measure answer quality systematically and reduce the incidence of fabricated responses.

Programme

What you will learn

Part 1 · 2 modules

Foundations

1.1Why retrieval matters
  1. (a)The limits of a model's own memory
  2. (b)Grounding answers in source documents
  3. (c)Retrieval compared with fine-tuning
1.2Preparing knowledge
  1. (a)Document cleaning and structure
  2. (b)Chunking strategies
  3. (c)Embeddings and vector databases

Part 2 · 2 modules

Building the assistant

2.1Retrieval
  1. (a)Semantic, keyword and hybrid search
  2. (b)Re-ranking results
  3. (c)Filtering by metadata and permissions
2.2Generation
  1. (a)Prompting the model to answer from context
  2. (b)Citations and source links
  3. (c)Saying so when the answer is not found

Part 3 · 2 modules

Quality and upkeep

3.1Measuring quality
  1. (a)Test questions and expected answers
  2. (b)Faithfulness and relevance scoring
  3. (c)Collecting user feedback
3.2Keeping it current
  1. (a)Updating the index as documents change
  2. (b)Access control and confidential material
  3. (c)Version history for regulated content

Who should attend

  • Knowledge managers and policy owners
  • Developers building internal assistants
  • Training providers digitising courseware and policies
  • Compliance teams handling large document sets

In-house delivery

Tailored for your organisation.

Delivered at your offices or conducted virtually, each run is customised around your active course portfolio and internal operating procedures.

Book a consultation All workshops