Skip to main content
← Documentation

AI Workflows

AI Workflows

As an operator, you can monitor AI workflows as they process tickets, review proposed changes, and control workflow execution. This page covers the operator-facing workflow tools.

For a general overview of how AI workflows operate, see the AI Workflows overview.

Monitoring active workflows

When an AI workflow is running on a ticket, you can follow its progress in real time from the ticket detail view or the Workflows section.

The workflow panel shows:

  • Current step -- What the AI is doing right now (for example, analysing the codebase, preparing a fix, creating a pull request)
  • Step history -- A log of completed steps with timestamps
  • Live output -- Streaming text output from the AI as it works

Workflow modes

Each ticket can be processed in one of four modes. The mode is typically set by the portal's configuration or chosen when the workflow is triggered.

ModeWhat happens
Full processingThe AI analyses the issue, investigates the codebase, proposes a fix, and creates a pull request
Analysis onlyThe AI investigates and posts findings but does not create code changes
Quick fixThe AI attempts a targeted fix without a full investigation -- suited for straightforward issues
Human onlyNo AI processing -- the ticket is handled entirely by your team

For more on how each mode works, see Workflow Modes.

Approving or rejecting changes

When a workflow reaches a step that requires human judgement -- such as approving a proposed code fix or pull request -- it pauses and creates an approval request.

To review an approval request:

  1. Open the ticket or navigate to the approval from the Workflows section.
  2. Review the proposed changes. The workflow shows what files were modified and a summary of the changes.
  3. Choose an action:
    • Approve -- The workflow continues with the proposed changes.
    • Reject -- The workflow stops and your feedback is recorded. You can add a comment explaining why the changes were rejected.

Approval requests expire after 24 hours if no action is taken.

Retrying a workflow

If a workflow fails or produces unsatisfactory results, you can retry it:

  1. Open the ticket with the failed workflow.
  2. Click Retry Workflow.
  3. The workflow restarts from the beginning with a fresh run.

Retrying is useful when external conditions have changed (for example, a dependency was updated or additional context has been added to the ticket).

Cancelling a workflow

To stop a running workflow:

  1. Open the ticket with the active workflow.
  2. Click Cancel Workflow.
  3. The workflow terminates and its status is updated to cancelled.

Cancelling a workflow does not revert any changes that were already made (such as branches or draft pull requests). You may need to clean up those artefacts manually.

Workflow history

Each ticket maintains a history of all workflow runs. You can review past runs to see:

  • Which mode was used
  • What steps were completed
  • Whether the run succeeded, failed, or was cancelled
  • Any approval decisions and comments