Workshops: AI Automation & Agents
Workshop: Build a RAG Knowledge Assistant
In this workshop your team builds an AI assistant that answers questions reliably from your own documents and cites its sources. You understand every building block of a RAG architecture and learn to control answer quality, permissions and data protection.
Who is this workshop for?
- IT and data teams making internal knowledge accessible with AI
- Knowledge management, quality and documentation owners
- Technical product managers for AI projects
- Engineers working with LLM applications for the first time
- Innovation teams planning a RAG pilot
What you will take away
- You explain the RAG pipeline from document preparation to answers with citations.
- You prepare documents and choose chunking strategies for manuals, contracts or FAQs.
- You set up embeddings and a vector database and combine semantic with keyword search.
- You implement document-level permissions so nobody gets answers from areas they cannot access.
- You measure answer quality with a test set and improve it deliberately.
- You assess build options: off-the-shelf, no-code with n8n, or custom development.
Agenda
Module 1
Day 1: Foundations and architecture
- Why language models need your knowledge
- RAG vs. project knowledge in ChatGPT, Claude or Copilot
- Building blocks and data flow
Module 2
Day 1: Preparing data
- Ingesting PDFs, Word, Confluence, SharePoint
- Chunking, metadata and updates
- Exercise: analyse your document landscape
Module 3
Day 1: Retrieval and generation
- Embeddings and vector databases overview
- Hybrid search and re-ranking
- Prompts that require citations and admit not knowing
Module 4
Day 2: Hands-on prototype
- Prototype with n8n or Python and a model of your choice
- Connecting Claude, OpenAI or Mistral
- Interface in Teams or on the web
Module 5
Day 2: Quality, security, operations
- Test questions, scoring and typical failure patterns
- Permissions and prompt injection via documents
- EU hosting, costs and operating model
When a custom RAG assistant pays off
For small document collections, the project and knowledge features in ChatGPT, Claude or Microsoft 365 Copilot are often enough. A custom RAG assistant pays off when knowledge is spread across many systems, permissions must be fine-grained or answers must be traceable to sources.
We help you draw that line for your case before effort goes into a custom build.
Typical use cases include internal policies and manuals, technical documentation for service and sales, product knowledge for new hires or tender documents. They all share one thing: the knowledge exists but is hard to find.
The most common mistakes
Most RAG projects fail not because of the model but because of the data and missing quality measurement.
- Outdated or contradictory documents in the index
- Chunks that are too large or too small
- No test questions, so no measurable improvement
- Permissions considered only after the fact
- Answers without citations that nobody trusts
Data protection and security
A knowledge assistant often concentrates confidential content. We cover EU hosting, data processing agreements, logging and whether local models make sense for certain data. Documents can also contain hidden instructions – a risk we demonstrate and mitigate in practice.
Frequently asked questions
Do we need engineers for the workshop?
Technical basics help for the prototype. Mixed teams from IT and business work best, because good answers need both perspectives.
Can we use real documents?
Yes, if they are cleared for the exercise. Alternatively we use anonymised or public documents from your field.
Will you build the assistant for us afterwards?
On request, our AI Solutions Architects and AI Code Engineers take the prototype to production or support your team in doing so.
Workshop
Get in touch
Want your company knowledge to be one question away?
Tell us a little about your team – we will get back to you within 24 hours with a proposal for content, format and dates.
Prefer to talk?
More workshops
- Workshops: AI Automation & AgentsWorkshop: AI Chatbot for Website & Customer Service
- Workshops: AI Automation & AgentsWorkshop: AI Document Processing for Invoices & Contracts
- Workshops: AI Automation & AgentsAI for Excel & Data Analysis: ChatGPT, Claude & Copilot
- AI Development Services24/7 Support & Maintenance for AI-Built Apps
