Skip to main content
Custory AI works inside the same workspace surfaces your team already uses. Use it when you want AI help to stay attached to a journey, item, automation, or chat instead of living in a separate prompt history.

​
What this means in Custory

Custory represents AI as a real workspace member named Custory AI Bot. That affects how AI work behaves:
  • AI responses can appear in the same item comments and workspace flows as the rest of the team.
  • AI requests can run asynchronously instead of behaving like instant autocomplete.
  • The request, result, and follow-up stay tied to the workspace.
Custory AI is not a billable seat. You also cannot change its role, remove it from the workspace, make it the owner, or have it leave the workspace.

​
Where you can use it

​
What happens when you mention AI on an item

The item-comment path is the most direct way to use AI as a workspace member.
  1. Add a comment on a journey-attached item and mention @Custory AI.
  2. Custory creates an async agent run for that request.
  3. The run uses the current comment plus the recent comment history on that item as context.
  4. When the run finishes, Custory posts the result back to the same item as a new comment from Custory AI Bot.
  5. If the run fails, Custory posts an error comment instead of failing silently.
This path requires normal comment permissions. In practice, that means you need edit access on the journey item because viewers cannot add comments.

​
How chat and memory behave

In-app AI chat is workspace-scoped. If the chat is tied to a journey, Custory can also use that journey context. Workspace memory is on by default for chat sessions. You can turn Memory off for a specific chat when you want a narrower answer from only the current prompt. Turning memory off skips retrieval for that chat. Useful outcomes from the chat may still be learned into workspace memory afterward. Read Workspace memory for the retrieval model and source-status behavior.

​
What stays visible and reviewable

Custory does not treat AI work as invisible background activity.
  • Item-mention requests create tracked async runs instead of one-off hidden prompts.
  • Item replies appear in the same comment stream as the original request.
  • Completed agent runs are ingested into workspace memory for later retrieval.
  • When chat creates reviewable items, they can stay pending until you apply or reject them.
Use this model when the team needs speed without losing the reason behind a suggestion or update.