Troubleshooting Issues with Autopilot AI Agents

If your Autopilot AI Agents are not performing actions as expected, this guide will help you identify and resolve common causes. Most issues occur due to configuration gaps, permission restrictions, or environmental factors like category or integration mismatches.

1. Verify the Agent’s Status

Start by confirming that the AI Agent is currently Enabled.

• Navigate to Autopilot → Hired Agents.
   
• Check the green status indicator on the agent card.
   
  • Green = Enabled
     
  • Grey = Disabled
     

If it’s disabled, open the agent and click the Enabled/Disabled toggle to activate it again.

2. Check the Agent’s Category Association

AI Agents are deployed for specific categories during configuration.

If a ticket is moved to a category not associated with that agent, the agent will no longer take action on that ticket. And all previously queued actions will fail.

To verify:

1. Go to Autopilot → Hired Agents.
    
2. Click Configure on the affected agent.
    
3. Under Category Association, ensure the target category is listed.
    
4. Re-deploy if changes were made.

3. Check for Missing Dependencies

AI Agents depend on a set of supporting configurations within your Help Desk and within their own onboarding setup. If any of these dependencies are modified or deleted after deployment, the AI Agent may stop functioning and be disabled.

A. Help Desk - Specific Action Dependencies

These are native elements within your HappyFox Help Desk that the AI Agent interacts with when performing actions.

Common examples include:

• Ticket Custom Fields (e.g., “Order ID”, “Issue Category”)
   
• Contact Custom Fields (e.g., “Customer Tier”, “Account Type”)
   
• SLAs referenced for timing or escalation logic
   
• Other Help Desk entities such as Statuses, Priorities, or Tags that the agent might attempt to modify
   

If any of these are deleted, or removed from a category after the agent was deployed, actions depending on them will fail.

• In the Activity Page, the Execution Status may show as Failed or the entry may not appear at all.
   
• The AI Agent will not automatically recover when the dependency is restored.
   

To fix:

1. Recreate or restore the missing Help Desk element (field, SLA, etc.).
    
2. Open the affected AI Agent in Autopilot → Hired Agents → Configure.
    
3. Review its setup to ensure dependencies are correctly linked.
    
4. Click Deploy [Agent Name] again to reapply the configuration.

B. Business Configurable Dependencies

These are custom parameters defined during the AI Agent’s onboarding flow. They guide how the agent evaluates information and makes decisions.

Examples include:

• Compliance instructions that restrict how content is generated or flagged
   
• Return policy windows used by refund or order dispute-related agents
   
• Threshold values or numeric parameters controlling when an agent triggers an action
   

If these are removed, left blank, or set incorrectly, the agent may misinterpret conditions or not execute at all.

To fix:

1. Open the AI Agent’s configuration from Hired Agents → Configure.
    
2. Revisit any text, numeric, or dropdown fields under the configurable section.
    
3. Update the values based on your current business rules.
    
4. Redeploy the agent to confirm the changes.
    

How to Identify These Issues

• In most cases, the AI Agent Activity table will show entries with a status of Failed or No Action Taken.
   
• No error details are shown in the interface.
   
• If an expected entry does not appear, assume a dependency issue is preventing execution.

4. Review the Agent’s Activity History

Navigate to the AI Agent Activity Page to confirm if the agent attempted execution:

• Go to Autopilot → Hired Agents → [Agent Name] → Activity.
   
• Check the Execution Status column.
   
  • Completed – The agent performed the action successfully.
     
  • Pending – The action is queued for execution (allow 1–2 minutes).
     
  • Waiting for Approval – In Supervised Mode, awaiting manual review inside the ticket.
     
  • Cancelled – Rejected or invalidated due to configuration issues.
     

If actions consistently show as Failed or Cancelled, re-verify configurations and integrations.

5. Check for Integration-Related Issues

If your AI Agent is linked to an external integration (e.g., Shopify, Stripe, etc), ensure that:

• The connected integration is still active under Manage → Integrations.
   
• The integration token or authorization has not expired.
   
• The integration is functioning normally (try a manual sync or test event).
   

If an integration is disabled, all agents dependent on it will stop functioning until reconnected.

6. Review Smart Rules and Automations

Smart Rules, Workflows, or other automation mechanisms may override or conflict with AI Agent actions.

For example:

• A Smart Rule may reassign a ticket after an AI Agent adds a tag.
   
• An automation might change a status immediately after an AI action, reversing its effect.
   

To avoid conflicts:

• Review your Smart Rules under Manage → Smart Rules.
   
• Check the order and trigger conditions to ensure they don’t execute in response to AI Agent actions unless intentional.

7. Confirm Permissions

Only users with the Manage Autopilot permission can manage, configure, or view AI Agent activity.

If other team members report missing access or see only a read-only view of the Marketplace:

• Go to Manage → Roles and Permissions.
   
• Enable the Manage Autopilot permission for their role.

8. Supervised Mode Approvals

If your AI Agent is running in Supervised Mode, pending actions must be manually approved:

• Go to the relevant ticket.
   
• Locate the AI Agent Pending Approval banner or update in the ticket timeline.
   
• Approve or reject the suggested action.
   

Pending items remain in the agent’s activity view as “Waiting for approval” until a decision is made.

9. Re-Deploy if All Else Fails

If configuration changes or category updates were made recently, re-deploying ensures that the latest setup is applied.

To redeploy:

1. Open the agent from Hired Agents.
    
2. Click Configure → Deploy [Agent Name].
    
3. Confirm deployment and wait for completion.

10. Check for Autopilot Credit Exhaustion

If your AI agents suddenly stop functioning without any configuration or permission changes, it could be due to running out of Autopilot usage credits.

Symptoms

• Agents show as “Enabled” but do not perform any actions.
   
• No new entries appear in the AI Agent Activity page.
   
• “Failed” or “No execution” status appears in the Activity table without further error details.
   

What to Check

1. Navigate to Autopilot → Usage Overview (if available) or contact your account administrator.
    
2. Verify if you have enough credits for your AI agents to operate.
    
3. If credits are exhausted, agents will automatically pause execution until the next billing cycle or top-up.
    

Fix

• Admins can choose to purchase additional Autopilot credits to resume agent activity.
   

Tip: Autopilot will not disable your agents - it only halts action executions. The AI agents remain visible and configured but idle until credits are added.

11. Still Not Working?

If your AI Agent remains inactive after checking all the above:

• Temporarily disable and re-enable it.
   
• Verify that Autopilot is enabled at the account level.
   
• Contact [email protected] with details from your Activity page for further help.