Adding a knowledge source does not automatically update your agent. You must trigger a training run to rebuild the agent's knowledge index. This page explains when to retrain, how to do it, and how to monitor progress.
Retrain your agent whenever:
You do not need to retrain when:
Training begins immediately. The agent status changes to Training while the process runs.
If you need to add multiple sources or delete several outdated ones, make all of those changes before clicking Train Agent. Each training run processes all sources together — batching your updates means one training cycle instead of several.
Training status is shown in the agent header and the Knowledge Management panel:
| Status | Meaning |
|---|---|
| Trained | Training complete — agent is live and using the latest sources |
| Training | Retraining in progress — wait before testing or deploying |
| Issues | One or more sources failed to train — see Knowledge Management for details |
Training typically completes within a few minutes. Large knowledge bases or many simultaneous sources may take longer.
Your agent remains available to users during retraining, but it continues to use the previous knowledge index until the new training completes. Avoid deploying major knowledge changes during high-traffic periods.
Start building intelligent AI agents to engage with customers.
