When a Publish Fails

If a publish encounters an error, the publishing modal shows a Failed state. Here's what to do when this happens.

What the failed state means

A failed publish means the AI configuration changes you submitted were not successfully applied to the live system. The AI continues using the last successfully published configuration — your new changes are not live.

Your changes are still saved in the dashboard as a draft. You haven't lost anything; the publish just didn't complete.

How you're notified

In the publishing modal: The modal shows Failed with an error message if one is available. The message may describe what went wrong.

In the Notification Center: If you dismissed the modal, a System category notification is sent when the publish fails. Go to Dashboard → Notifications → System to see it.

Common reasons a publish fails

  • Temporary service disruption: Speako's background processing service may have been momentarily unavailable. This is the most common reason for a failed publish.
  • Configuration error: In some cases, a specific part of the configuration (e.g. an invalid greeting format) may cause the job to fail.
  • Timeout with no completion: Very rarely, a job may time out without completing.

What to do

Step 1: Try publishing again. Most publish failures are transient. Wait 2–3 minutes, then click Publish again for the same configuration. The second attempt usually succeeds.

Step 2: Check for a service maintenance notice. If multiple publish attempts fail, check your Notification Center for any maintenance or service disruption notices in the Maintenance tab.

Step 3: Contact support if the issue persists. If several attempts fail over a period of time, reach out to Speako support. Provide the approximate time of the failed publish attempts and what type of changes you were publishing (knowledge, greetings, tools, etc.).

⚠️ Important: A failed publish means your most recent changes are not live on the AI. If you've made time-sensitive updates (e.g. holiday closures, greeting changes), confirm the publish succeeds before assuming callers will experience the updated version.

💡 Tip: After a failed publish and a successful retry, make a test call to confirm the expected changes are reflected. This is the most reliable way to verify the publish succeeded.