Learning from past tickets

TR-AI learns from tickets your team resolved — only human answers, never its own.

All plans (history window: 6, 12 or 24 months) Open in TR-AI4 min read

When you connect a help desk, TR-AI imports solved tickets from your history window and keeps learning from newly solved ones every few minutes. Each past ticket becomes a question-and-answer pair; a new ticket is matched against the old question on its own and against the full conversation, so repeat questions are recognised even when phrased differently.

Only replies written by people count. Replies TR-AI posted (and anything created through the API or by automations) are skipped.
If a person replied after TR-AI on a ticket, only the person’s reply is learned.
Agents’ internal notes are used as background for your team’s summaries, never quoted to customers.

1
Re-import when needed
Overview → Setup & connections → Past tickets → Re-import replaces the stored history with a fresh import (for example after cleaning up macros).
2
Check what was skipped
The Past tickets card shows how many solved tickets were skipped because TR-AI or an integration answered them.

Tips & tricks

Good past answers are worth as much as articles: encourage agents to write complete answers rather than “see article X”.
If a past answer is out of date, fix the article — current knowledge-base content wins when they disagree, and conflicts are flagged.

PreviousLearning your product (domain learning)NextFree site search and paid web search