Knowledge Management

Add and organize knowledge for your Assistant

Your AI Assistant primarily learns from:

The documents you upload, like PDFs, CSV files, etc., and the URLs you provide.
Notion pages and SharePoint documents if you connect your Notion/SharePoint account to Atomicwork.
FAQs which can be saved as Verified Answers.
Conversations in configured public Slack channels, limited to the users or user segments selected by admins.

To configure Slack channel learning, see Learn from Slack channels.

These are all the supported file formats: DOC/DOCX, PPT/PPTX, XLS/XLSX, PDF, ASPX, CSV. It can also learn from the transcript in YouTube links!

Add knowledge

Navigate to Settings > Your Workspace> Knowledge.
Click on Add.
Enter a name and define the segment for it.
Click on Add to publish the content. You can choose to publish it later whenever you're ready for the Assistant to learn from it.

Add a document

Choose the folder you want to add a document to.
You can upload files with any of the following extensions: DOC/DOCX, PPT/PPTX, XLS/XLSX, PDF, ASPX, CSV, or TXT. You can also add a URL for the Assistant to find and learn from or sync documents from Notion or SharePoint.
- You can choose the audience for the document. This means, the Assistant will surface the answers from these documents only to relevant segments. For content in SharePoint, Atomicwork inherits permissions directly from SharePoint.
- You can also restrict end users from downloading the source documents.
- If you add a URL or a Notion/SharePoint document, the Assistant will relearn from this content every few hours to stay updated.
  - NOTE: The advantage of adding a URL is that it will stay up to date on even small content updates without you having to upload the document again every single time.
Once you add a document, Atomicwork will mark the status as “Processing” till the Assistant is done learning from the content.
If it fails to do so, we’ll notify you so you can optimize your document better before reloading it for the Assistant to learn from.

We use OpenAI to help the Assistant be a conversational star and we do not give it any of your documents for training purposes so you can rest assured on that front.

Set audience for knowledge

When you add knowledge, you can also specify which employee segments have access to the content. When you specify a segment, the Assistant will provide answers from the content only to the employees in the segment.

When allowing Atom to learn from Slack channels, admins can choose which employee segments receive answers learned from a particular channel. Note: Users in the selected audience segment can receive those answers even if they are not members of the Slack channel where the source message was posted.

You can set the segment for the all the content you upload and any documents you bring in via Notion, but you cannot set the segment for SharePoint documents in Atomicwork. SharePoint docs retain the permissions that the creator set in SharePoint or through Azure Active Directory.

Navigate to Settings > Your workspace.
Go to Knowledge > Document name > Edit.

If an employee asks questions about the content in a public channel, the Assistant will DM them the answers to preserve confidentiality and avoid information burden.

Restrict source download

Go to Settings > Your workspace > Knowledge > Name > Document name > Edit > Segment Settings. Toggle Restrict download.
If you enable this setting, your employees will not be able to view or download this document.

FAQs

Are there any restrictions to the types of content the Assistant can learn from? The Assistant cannot read images or code in your documents/Notion pages. This means that it will not be able to parse text inside images or “learn” from code.
How does it handle abbreviations? If it’s an abbreviation, consider adding an expansion for the abbreviation at least once in the document so the Assistant knows what it means. For example, if you refer to Macbook Pro devices as MBP in your device policy document, add at least one Macbook Pro (MBP) reference for the Assistant to learn from.

Add a document

Choose the folder you want to add a document to.

You can upload files with any of the following extensions: DOC/DOCX, PPT/PPTX, XLS/XLSX, PDF, ASPX, CSV, or TXT. You can also add a URL for the Assistant to find and learn from or sync documents from Notion or SharePoint.

You can choose the audience for the document. This means, the Assistant will surface the answers from these documents only to relevant segments. For content in SharePoint, Atomicwork inherits permissions directly from SharePoint.
You can also restrict end users from downloading the source documents.
If you add a URL or a Notion/SharePoint document, the Assistant will relearn from this content every few hours to stay updated.
- NOTE: The advantage of adding a URL is that it will stay up to date on even small content updates without you having to upload the document again every single time.

Once you add a document, Atomicwork will mark the status as “Processing” till the Assistant is done learning from the content.

If it fails to do so, we’ll notify you so you can optimize your document better before reloading it for the Assistant to learn from.

We use OpenAI to help the Assistant be a conversational star and we do not give it any of your documents for training purposes so you can rest assured on that front.

Set audience for knowledge

Navigate to Settings > Your workspace.

Go to Knowledge > Document name > Edit.

If an employee asks questions about the content in a public channel, the Assistant will DM them the answers to preserve confidentiality and avoid information burden.

FAQs

Are there any restrictions to the types of content the Assistant can learn from? The Assistant cannot read images or code in your documents/Notion pages. This means that it will not be able to parse text inside images or “learn” from code.

How does it handle abbreviations? If it’s an abbreviation, consider adding an expansion for the abbreviation at least once in the document so the Assistant knows what it means. For example, if you refer to Macbook Pro devices as MBP in your device policy document, add at least one Macbook Pro (MBP) reference for the Assistant to learn from.