# Resolving failures

It is important to transition your business reliance on automation gradually, testing new automations for a couple weeks for success before handing the wheel entirely to our system. Despite our efforts to provide you with maximum robustness here, new popups or unexpected data can still cause automations to fail. Should this happen, here's how you can go about resolving failures in your automations.

# Identifying where an automation is failing

This is often the hardest part of addressing failed automations. You can significantly reduce the time this process takes by ensuring logs contain a detailed view of what the automation is doing at every step.

  • Navigate to the job details page for the failed automation, and click "View Logs" for the most recent failed run. This should give you a general idea of where your automation failed.

  • Review any other artifacts generated by your automation in our system (Data Tables, Files, etc.).

  • Log into the Agent Server running your automation, and monitor the desktop to review the automation visually. If you have multiple Agent Servers, and aren't sure which one your automation will be routed to, try marking all other servers as "disabled" where possible.

# Deploying a fix

Once you know where an automation has failed, edit your model with a fix and, if using a Python-based model, zip your files according to the directory structure provided on the "Python Models" page.

Click the corresponding "Actions" column button for your failing model in the "Models" page, click "Edit", confirm the type of your automation is set to the type of model you wish to upload, and click "Upload Automation Package ZIP file".

Once you select your new zipped code, click Update, and try running your job again. Your changes should be applied. If they weren't, you're likely running into an updating bug on our end. For the time being, we recommend uploading your fix as a new model, and creating a new job that uses the new model.