Asking Questions

Asking your first question

The AI Knowledge Agent accepts natural language questions — write the way you would ask a colleague. There is no special syntax or query language required.

You can ask questions through any supported interface:

• Slack — DM the Kowiki bot or @mention it in a channel
• Microsoft Teams — Chat with the Kowiki bot or invoke it in a channel
• Web wiki — Use the chat widget on your Kowiki site

Type your question, send it, and the agent returns an answer within a few seconds along with citations to the source pages.

Writing effective questions

Specific, focused questions produce the best results. Give the agent enough context to narrow down the relevant content.

Examples of effective questions

• "What is our refund policy for annual subscriptions?"
• "How do I configure SSO with Okta for our staging environment?"
• "What are the required fields when creating a new customer record in the CRM?"
• "Who is responsible for approving travel requests over $5,000?"

Examples of vague questions that produce weaker answers

• "Refunds" — too broad; the agent does not know what aspect you need
• "How does it work?" — no context about what "it" refers to
• "Tell me everything about onboarding" — too wide; ask about a specific step instead

Tips for better questions

• Include the specific topic or system you are asking about.
• Mention the scenario or use case when relevant.
• Ask one question at a time for the clearest answers.
• Use terminology that matches your documentation (product names, internal terms).

Follow-up questions

The agent supports multi-turn conversations. Ask follow-up questions in the same thread or conversation, and the agent uses the prior context to interpret your next message.

For example:

1. You: "What is our policy on remote work?"
2. Agent: (answers with citations)
3. You: "Does that apply to contractors as well?"
4. Agent: (answers using context from the previous exchange)

The agent retains conversation history within a single thread. Starting a new thread or conversation resets the context.

When the agent cannot answer

If the agent cannot find relevant content in your knowledge base, it tells you directly rather than fabricating an answer. You may see a response like:

"I couldn't find information about that topic in your knowledge base."

When this happens:

• Rephrase your question using different terminology.
• Check whether the relevant documentation exists in your wiki.
• Verify that you have access to the section where the information lives. The agent only searches content you are permitted to view. See Section-Level Permissions for details.

Conversation history

Each conversation thread maintains its own history. The agent references earlier messages in the thread to provide coherent follow-up answers. History is not shared across separate threads or between different users.

To start fresh, begin a new conversation or thread. This is useful when you switch to an unrelated topic.