Knowledge Base for Sales and Engineering: Which Content to Prioritise First

A knowledge base does not need to begin with all of the company's knowledge. For an initial, focused scope, a different question is more useful: Which content should answer specific questions from sales and engineering, and who verifies that it may be used?

The following Trixner Working Model is an editorial checklist, not a universal standard. It contains exactly six review areas. Each one provides questions for internal clarification, not a prescribed solution.

Retrieval-Augmented Generation (RAG) is an approach in which a language model retrieves information from an external knowledge source to generate an answer. The original RAG paper describes this as combining a generative model with retrievable, non-parametric knowledge.

Metadata is descriptive information about data or digital content. For an internal inventory, companies can define the labels they want to use, such as source, responsible role or internal approval status.

Begin by selecting a clearly bounded knowledge area, such as one product group, a recurring type of enquiry or a defined technical service. Then review the available content across the six areas.

Practical Tip: As a starting point, collect five to ten real questions from sales and engineering. This number is a working suggestion, not a universal minimum. For each question, note which internal source should currently be used to answer it.

Content is prioritised when its relevance to the selected knowledge area can be described clearly. Instead of importing entire folders indiscriminately, assign each document to a specific question or task.

• Which question should this content answer?
• Which product group, service or type of enquiry is it intended for?
• Which parts are not required for the selected initial scope?
• Which questions remain open despite this content?

For each document, define internally how its intended status for use will be identified. The decisive factor is not a particular technical field, but a rule the company understands.

• Which version should be used in the selected area?
• How do employees identify the internally approved version?
• How should drafts, superseded versions and valid content be distinguished?
• Who decides when document versions conflict?

The origin of content should remain visible in the inventory. The company can define appropriate metadata for this purpose. The exact selection depends on its own process.

• Where does the content come from?
• Which product, customer case or internal process does it relate to?
• Which label should be visible in an answer or review?
• Which terms or abbreviations need to be explained internally?

As a working rule, assign a subject-matter owner to each knowledge area. This role does not have to maintain all content personally, but should be able to decide how follow-up questions and unclear versions are handled.

• Who can review the content from a subject-matter perspective?
• Who may approve a version for its intended purpose?
• To whom are unclear or contradictory statements returned?
• How should cover for the role be arranged?

Document the required access as a business requirement before selecting a technical solution. This keeps decisions about roles and content independent of a particular platform.

• Which roles should be allowed to view the content?
• Are any parts intended only for internal use or for specific tasks?
• Which information may appear in a draft answer?
• What should happen when permission is missing?

Agree a dedicated review process for the focused initial set. The checklist does not prescribe a universal interval. The company defines the date, trigger and responsible role to suit the selected content.

• When should the content be reviewed again?
• Which change should trigger an earlier review?
• How is feedback from sales and engineering incorporated?
• What should happen to content that is no longer intended for use?

Practical Tip: For the workshop, use a simple table with the columns “Question”, “Source”, “Approval status”, “Responsible role”, “Access requirement” and “Next review”. These columns are a suggestion and can be adapted to the company's workflow.

After the inventory, the company can define its own sequence. One possible approach begins with content assigned to a clear initial scope, with an internally confirmed status and a responsible role. Documents with open questions remain visible on the clarification list.

The initial result is not a finished knowledge base. It is a focused, transparently described selection that can be reviewed by specialists and then assessed technically.

The first step does not require every document to be imported. A focused set whose purpose, approval status, origin, ownership, access requirements and maintenance can be clarified internally is more useful.

The technical decision between conventional search and RAG deliberately comes later. To structure your initial scope around specific questions, see RAG for Companies for guidance on the appropriate next step.

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Digital information management factsheet: What is metadata?