Automation Failures and Recovery

Even well-built automations fail sometimes. External APIs change, credentials expire, and unexpected data shapes arrive. This guide covers how Chase Agents handles failures, the auto-fix recovery process, and what to do when recovery does not work.

What Happens When a Step Fails

When a step throws an error, Chase Agents records the failure in Workflow History with the full error details. What happens next depends on the automation's error handling mode. In ai mode, the platform immediately attempts to fix the failing step using the AI auto-fix engine. In any mode, the run stops immediately and the failure is logged without any repair attempt.

The Auto-Fix Recovery Process

When auto-fix engages, it analyzes the failing step's definition, its inputs, and the error message. It generates a repaired version of the step and re-executes it. If the repaired step succeeds, the automation continues from that point. The repair is saved to the automation definition so future runs use the fixed version. If the repair fails, the run is marked as failed and no further auto-fix attempts are made for that run.

When Auto-Fix Keeps Failing

If auto-fix fails repeatedly on the same step, the underlying issue is likely beyond what the AI can repair automatically. Common reasons include: the automation is fundamentally trying to do something the connection does not support; the external service has undergone a major API version change; the automation's core logic has a design flaw that cannot be patched at the step level. In these cases, manual intervention is needed.

Disabling an Automation That Keeps Failing

If an automation is running on a schedule and failing every run, disable it while you investigate and fix the root cause. Running a failing automation repeatedly wastes action quota and generates noise in Workflow History. Disable the automation, analyze the Workflow History for the pattern of failures, identify the root cause, make the necessary repairs in the visual editor, and then re-enable it.

Post-Execution Auto-Fix

When post-execution auto-fix is enabled on an automation, the platform analyzes the entire failed run after it completes. Rather than just fixing the step in isolation, post-execution analysis can identify when a failure earlier in the run caused downstream failures, and repair the root cause step rather than a symptom step. This produces more comprehensive repairs than the in-run step-level auto-fix.

Reviewing the Auto-Fix Change

Whenever auto-fix modifies your automation, review the change before relying on it in production. Open the automation in the visual editor and inspect the repaired step. Verify that the repair is logically correct and not just syntactically valid. Auto-fix can produce repairs that make a step stop erroring without actually producing the right output. Check the outputs in Workflow History to confirm the repair is producing sensible results.

Setting Up Failure Notifications

For production-critical automations, set up failure notifications so you are alerted immediately when a run fails. You can do this by adding a notification step at the end of your automation that only runs if a prior critical step failed, using conditional logic to detect the failure state. Alternatively, build a separate monitoring automation that periodically queries Workflow History for failed runs and sends a summary to your team's Slack channel.