Your knowledge base ready for AI.
A knowledge base is ready for AI when every topic has one up-to-date, consistent version, the content is easy to find and someone is responsible for its maintenance. AI applications such as Agent Assist or a chatbot search this knowledge and base their answers on it. If your articles are contradictory, outdated or incomplete, the answers will be too.
- One up-to-date source per topic
- Clear ownership and management
- Test with real customer questions
5 min read · Interactive knowledge base check
A strong knowledge base is built step by step.
What return conditions apply now?
Bring sources together and remove duplicate or outdated versions. This makes it clear which agreement takes precedence.
Start with the customer question. First choose the knowledge you need for one application.
Who keeps this knowledge reliable?
Assign articles an owner and review date. Have important changes checked by a second person.
Make management part of the work. Knowledge requires a regular maintenance routine.
Does the knowledge provide sufficient guidance?
Test with frequently asked questions as well as difficult exceptions. Use feedback and unanswered questions for improvement.
Keep learning from usage. Fill in what is missing and check again.
Illustration with fictional content. Not a live product view.
Why knowledge determines how well AI answers
Most AI applications for customer service use retrieval-augmented generation (RAG). The AI first searches your own knowledge for passages that match the question. It then formulates an answer based on those passages.
A missing exception results in no answer, or an answer that is too general. If two versions of the same policy exist, the AI may find the outdated version. And vague wording, unexplained abbreviations or simply “see attachment” do not provide enough guidance.
Is your knowledge base ready for AI?
Answer nine short questions about your first application. Discover what is already in place and where it is best to start.
This is your starting point.
There are still important areas to work on. Start with a small domain and tackle the actions below in order.
You already have some components in place. Address the open points in a focused way before expanding the application further.
Based on your answers, all applicable components are in place. Now test this with real customer questions and continue maintaining the knowledge.
Multilingual support is not applicable and is not included in the count.
Start with these improvement actions.
First, address what has not yet been arranged or is still unknown, followed by what is partially in place. If the answers are tied, we follow the order of the step-by-step plan.
Define customer questions
Choose one domain for your first application and collect the most common customer questions.
Determine authoritative sources
Create an overview of your sources and designate one authoritative source for each topic.
Clear out outdated knowledge
Archive old versions and merge overlapping articles within your chosen domain.
Clarify articles
Put the answer at the top and explicitly describe conditions, exceptions and next steps.
Build work instructions
Work out one frequently used process in numbered steps, with one action and clear decision points per step.
Assign ownership
Assign an owner and review date to each article. Have important changes checked by a second person.
Manage access
Define which information remains internal and which knowledge employees, specialists and customers may use.
Check languages
Check the most important articles in every required language and record which translations have been reviewed.
Test and improve
Create a test set with real customer questions and exceptions. Assess both answers and sources, and schedule a regular feedback session.
Take the step towards a targeted practical test.
Choose one application, test exceptions as well and assess whether answers are based on the right sources. Record who will follow up on feedback and when you will check again.
View and change your nine answers
- 01 · Define customer questionsIn placePartially in placeNot yet / I don't know
- 02 · Determine authoritative sourcesIn placePartially in placeNot yet / I don't know
- 03 · Clear out outdated knowledgeIn placePartially in placeNot yet / I don't know
- 04 · Clarify articlesIn placePartially in placeNot yet / I don't know
- 05 · Build work instructionsIn placePartially in placeNot yet / I don't know
- 06 · Assign ownershipIn placePartially in placeNot yet / I don't know
- 07 · Manage accessIn placePartially in placeNot yet / I don't know
- 08 · Check languagesIn placePartially in placeNot yet / I don't knowNot applicable
- 09 · Testing and improvingIn placePartly in placeNot yet / I don't know
This self-check is based on your answers. It is not a technical audit or a guarantee that your knowledge base is ready for AI.
Common mistakes
Trying to clean everything up at once
Start small with the domain of your first application and expand from there.
Treating knowledge management as a project
Knowledge becomes outdated. Without an owner and a set cadence, quality declines.
Looking only at good answers
Questions without answers are precisely what show where knowledge is missing.
Leaving exceptions in the team chat
What isn't in the knowledge base doesn't exist for the AI.
How Archie helps with AI-ready knowledge
Archie Knowledge is the AI knowledge base from Archie, the AI platform for customer service from VANAD Digital, part of VANAD Group.
Importing and synchronizing knowledge
With Knowledge Import & Sync, you can import knowledge once or keep externally managed articles synchronized from Salesforce Knowledge, Atlassian Confluence, Zendesk Guide, TOPdesk Knowledge Base or Contentful. You can see in advance what will be included.
Review before publication
The four-eyes principle and Article Pre-Check help assess articles. Scheduled articles remain hidden, and Archie only uses them after publication.
Search by meaning
Weak search results and expired articles are excluded before they can influence an answer.
Multilingual management
You manage knowledge by language, distinguishing between manual and automatic translations.
Your questions.
Clearly answered.
Learn more about preparing, managing and improving your knowledge base.
