Plan requirement
| Subscription | Any plan |
| Access | Admin |
Retention for AI agent conversations. A category that grows quickly and is easily overlooked.
Create one
- Open Admin Center, then Account, then Security, then Deletion schedules.
- Create a schedule for AI agent conversations.
- Set the period.
- Keep it long enough to be useful for improvement.
Why this category is different
Conversations that never became tickets still happened, still contain what customers wrote, and still accumulate. An AI agent handling hundreds a day builds a substantial store quite quickly.
They fall under the same obligations as anything else customers told you.
Do not set it too short
Transcripts are how you improve an AI agent. Six months of history lets you see whether a change helped; two weeks does not.
Retention and usefulness pull in opposite directions here, and the right period is a genuine trade-off.
What you lose
Once deleted, you cannot go back and read why the agent answered as it did. If somebody queries an answer given three months ago, the record has to still exist.
That argues for keeping them at least as long as your complaints window.
They contain personal data
Customers type order numbers, addresses and account details into chat exactly as they do into email. The conversation store is not a technical log.
Include it in a subject request
When somebody asks what you hold about them, AI conversations count. Worth remembering, because they sit outside the ticket list where people look first.
Comments
0 comments
Article is closed for comments.