Prepare your knowledge base
This article will help you prepare and curate your content that powers Josef Q.
Josef Q performs best when your knowledge base is complete, well-organised and curated around a clear purpose. The quality and structure of the content you provide has a direct impact on the quality of the answers Josef Q returns. Below we cover the file types Josef Q supports, how to get the best results from Word and PDF files, and how to request help when your content lives somewhere less straightforward.
Recommended content
Josef Q works best with content that is already organised as explicit, searchable knowledge. Good candidates include:
- Policies, procedures, handbooks, FAQs, governance and legal content
- Knowledge base and reference material with clear sections
Content to avoid
We do not recommend using training video transcripts. This content is difficult for Josef Q because transcripts often contain conversational or instructional language where key information is implied and spread through dialogue, rather than structured as explicit, searchable knowledge with clear topics and headings.
Focus over breadth
Josef Q performs best when content is curated around a clear purpose and a defined set of expected user questions. A smaller, focused collection of high-quality content will generally outperform a large repository of mixed or loosely related material.
- Start with content that directly supports your intended use case and user needs.
- Expand gradually as gaps are identified.
- Avoid adding large amounts of unrelated, duplicate, outdated, or low-value content, as this can introduce noise and reduce answer quality.
File support
Josef Q supports DOCX and PDF documents as standard content sources.
💡 Tip: If you have a choice between formats, a well-structured DOCX or a PDF exported from Word will work best with Josef Q. Read below for specific guidance.
Best practices for Word and PDF files
A little structure goes a long way. Following these practices when preparing your DOCX files helps Josef Q find and return the right information.
Do:
- Use clear section headings and a logical structure
- Keep each section focused on a single topic
- Keep related information together
- Write rules explicitly rather than relying on assumptions
- Use simple, clearly labelled tables where possible
- Include context such as thresholds, roles, combinations, and exceptions where relevant
Avoid where possible:
- Complex visual layouts
- Very large or heavily merged tables
- Critical information contained only in images or diagrams
- Splitting connected information across multiple documents
💡 Tip: Use Word's built-in heading styles (Heading 1, Heading 2, Normal, List) rather than styling text by font size alone. This is what allows Josef Q to understand your document's hierarchy.
For PDF files, check out our article explaining how to Optimise your PDFs for Josef Q.
How to request for help
Beyond DOCX and PDF, Josef can also support additional content sources, such as spreadsheets, information only available on websites, and other systems.
These sources often require discovery, extraction, transformation, and structuring work before they can be used effectively. For this reason, they are typically delivered as a separate Professional Services engagement.
If you have content like this that you'd like to bring into Josef Q, get in touch to scope the work:
- Contact your Josef account manager, or
- Email support@joseflegal.com
We'll work with you to understand the content, the effort involved, and the best way to prepare it for Josef Q.