How to build a private AI Team Wiki
Give employees one place to ask questions across approved internal documents and systems.
The Team Wiki uses the data sources of one chatbot and presents them on a separate subdomain. Employees ask a question and receive a direct answer from the indexed material.
Use a dedicated chatbot for the wiki. If you protect the wiki with a password, the regular website widget of that chatbot stops working and only the wiki remains available.
What you will have at the end
- A dedicated Team Wiki chatbot
- A configured subdomain, access mode and brand
- A real answer from an approved internal source
Before you start
- Standard plan or higher
- An approved internal source with a clear owner
- Premium or Enterprise for native Confluence and Notion connectors
- A decision between password, member login and public access
One chatbot supplies the knowledge
The Team Wiki reads the same websites, files, Confluence spaces and Notion databases as the selected chatbot. Branding and access settings change who can use it, not how retrieval works.
A separate wiki chatbot prevents internal access rules from affecting a public support widget. It also keeps sources, prompts and permissions easier to review.
01–09
Set it up step by step
Create a dedicated Team Wiki chatbot
Keep internal knowledge separate from your public widget.
Create a new chatbot such as “Internal Team Wiki”. Choose the model and system prompt for internal answers, then keep this assistant out of the public website.
A fast general model is usually the right starting point for an internal wiki. Better source structure often improves answers more than moving to a larger model.
Add approved internal knowledge
The wiki can only answer from the sources of this chatbot.
Open Data Sources and add maintained websites, files, Confluence spaces or Notion databases. Give each source a clear name and category.
For this test we uploaded a small internal document with the unique value NORDSTERN-42. A unique fact makes it easy to prove that retrieval works.
Open the Team Wiki settings
Read the access warning before creating the wiki.
Go to Integrations and select AI Team Wiki. The page shows the future subdomain, the current status and the Create Team Wiki button.
Live chat and human takeover are not available inside the Team Wiki. Plan a separate support path if employees still need a person.
Set title, subdomain and access mode
Choose an address and decide who may enter.
Enter the visible wiki title and an available subdomain. The availability check must be green before creation.
Public allows anyone with the link. Password uses one shared secret. Login Required allows the account owner and invited team members with individual accounts.
Understand what password protection changes
Password mode turns this chatbot into a wiki only assistant.
When Password is selected, the normal website widget for this chatbot no longer works. Website visitors would face the wiki password instead of the public chat.
Use a dedicated chatbot if you need both a public support widget and a private Team Wiki. This is not just a recommendation about organization; it prevents a real service interruption.
Finish branding and wiki options
Make the scope clear before employees start asking questions.
Upload an optional logo, choose the theme color and write a welcome message that says what the wiki contains. File Upload lets eligible users attach a file during a wiki conversation. Activate wiki immediately publishes the new subdomain as soon as you create it.
Keep the welcome text short. Tell users which topics are covered, how fresh the sources are and where they should verify policy critical answers.
Open the protected wiki
The access screen must appear before any internal content.
Open the generated subdomain in a private browser session. Password mode must show the Wiki Access dialog. Login Required must redirect to the member login instead.
Test both a correct and an incorrect credential. Do not rely only on a hidden link in the dashboard.
Ask a question with a known answer
Prove that the wiki retrieves the intended source.
We asked “What is the internal escalation code?” The real wiki answered NORDSTERN-42 from the uploaded internal document.
Repeat the test with a question whose answer is intentionally absent. The assistant should say that the information is missing instead of inventing a policy.
Review the active wiki in the dashboard
Keep ownership, access and availability visible.
The Team Wiki card shows its subdomain, creation date and active status. Use its actions to open, edit, disable or delete the wiki.
Disabling the wiki is the safest immediate response to a source or access incident. Deleting it should be reserved for permanent removal.
Example & result
See the practical test and its result
Every tutorial includes a fixed input, the expected outcome and a transparent record of what was actually verified locally.
Practical example: How to build a private AI Team Wiki
This exact scenario was completed with the temporary tutorial account.
Exact test input
Ask the private wiki: “What is the internal escalation code?”
Expected result
The password gate appears before the wiki and the answer contains NORDSTERN-42 from the approved internal document.
What was actually verified
The private subdomain required the configured password. After login, the real Team Wiki answered NORDSTERN-42 from the indexed internal source.
Tips & tricks
Make the setup reliable
Test with realistic examples, record your baseline and change one setting at a time. That makes real improvements visible.
Write documents for retrieval
Use clear headings, one topic per section, explicit dates and complete statements. Scanned pages, unexplained tables and repeated copies make answers less reliable.
Give every source an owner
Record who approves the content and when it must be reviewed. Remove expired policy versions instead of leaving the model to choose between them.
Use member login for accountable access
A shared password is quick, but individual accounts are easier to revoke and audit. Wiki only members can use the wiki without seeing the dashboard.
When something does not work
Troubleshooting
Check status, permissions and test data systematically before changing the model or prompt.
The expected option is missing
Confirm the account plan, feature permissions and selected chatbot. Paid or beta features can be hidden when prerequisites are not met.
The test result is inconsistent
Reset the test conversation, keep the input identical and change one setting at a time so the cause remains measurable.
Ready for a production-style test
Start with one department. Review failed questions, remove duplicate content and confirm source owners before adding more people or systems.