Proactive Alert Monitoring: LLM Model Latency

Resolution:

Here are some common practices & resources that may help to resolve issues related to this alert:

• Review any recent changes to the affected agents.

  • Changes such as switching to a different model, significantly longer instructions or prompts, or adding more actions and knowledge lookups per response can increase response time.

  • If latency increased right after a change was deployed, consider rolling it back to confirm.

• Confirm whether the slowness affects all conversations or only specific agents or channels, and share this detail in your support case — it helps speed up the investigation.

• If no recent changes were made on your side, consider this as an issue on the Salesforce side.

  • Elevated model latency is often caused by temporary conditions with the underlying AI service. You will be notified through your support case as the issue is investigated and when the service is restored.