Overview
The AI Activity Log is a centralized page in Messaging where your AI Agent's learnings are stored over time. Copilot or Autopilot picks up on a new learning from a guest reservation conversation when a Breezeway user replied to something that AI couldn't answer.
Within the Activity Log, you can reinforce a learning by adding it to Blueprint or a Property Profile so the rest of the team can take advantage of it, or discard a learning (by clicking Delete/Undo) if the AI shouldn't be using it going forward. It's the single surface where you can see what their AI has been picking up on and decide what to keep, where it came from and where there are opportunities to promote it to other parts of the product (Blueprint or Property Profile) to crystalize the knowledge further.
The AI Activity Log is available to Administrator and Supervisor roles when AI Agent is enabled (Copilot or Autopilot modes)
How the AI Activity Log Works
Learning Detection: When your AI Agent in Messaging is enabled (Copilot, or Autopilot) and a Breezeway user replies manually or edits a suggested AI response our AI analyzes that response to see if there's a valuable learning to capture so that it doesn’t repeat that same mistake twice.
Logging: If a learning is captured, it's added to the Activity Log as a new line item newest first, showing the AI's reasoning and the conversation context that triggered it.
Categorization: AI determines whether the learning is Company-level (general operational info), Property-level (specific to one or more properties), or Reservation-level (one-off context that shouldn't be applied broadly). The category determines what actions are available.
Note: Reservation level learnings won't have options to be added to blueprint or property profiles.
Suggested Actions: An admin or supervisor user in Breezeway can view the learning entry and has options to:
Add it to Blueprint or Property Profile: Makes it part of the standing knowledge base.
Undo/Delete the learning: Removes it from what the AI uses going forward.
Dismiss: Dismiss the suggestion and keep the learning but don't add it to any additional places in your Breezeway account.
Visual State: Once a “Suggested action” has been taken the line item moves from Suggested actions into “Learnings” with a checkmark success treatment if added to the knowledge base or red/discarded treatment if undone, along with a record of which user took the action.
Learning Types
When a learning occurs more than once, we make a suggestion for them to be promoted into the Blueprint as a source of truth for procedures and general guidelines, or the Property Profile for more property specific items.
Optimizing the AI Activity Log
Make it part of an admin's weekly routine: The value compounds when learnings are reviewed regularly. Encourage admins or supervisors to check the Activity Log weekly and action what's there.
Use the right destination: When adding a learning, choose Blueprint for general SOPs and Property Profile/custom fields for property-specific details. The right destination ensures the learning is applied broadly enough - but not too broadly.
Don't be afraid to Undo: If a learning was captured from a one-off scenario or contains information the AI shouldn't be using for some reason, undo/delete it. The Activity Log is meant to give you more transparency to what your AI is learning instead of being a black box.