You can see what every AI workflow is doing, from live progress to a clear record of what succeeded, failed, or was cancelled. This page collects where to look.
Following Progress
How you follow a run depends on how it was started:
|
How you ran it |
How you follow it |
|---|---|
|
Post-editing or quality assurance as a supplier service |
The sparkle icon in job lists and job details opens the AI Agent Work Status popup, which shows how far the run has got, for example Processing… 42%, and refreshes while you keep it open. The percentage counts the run's work batches and stays at 99% until the run itself reports completion. |
|
Machine translation during document preparation |
The Prepare files popup reports the phase with a percentage, for example Machine translation (de)… 42%. The line names the target language and starts again for each one. |
|
Machine translation from the Machine translate action |
The operation's progress window shows the run advancing, with a percentage where the AI reports one. |
|
Quality assurance from the Editor (Run QA) |
No percentage is shown: the QA run displays as in progress until it completes, and can be cancelled at any time. Findings appear in the QA panel when it finishes. |
|
A single segment in the Editor |
Nothing to track: the translation comes back directly. |
See AI-powered suppliers for the full monitoring tour: job list icons, tooltips, and the work status popup.
If a Run Fails
Every failure produces a readable message that includes two identifiers, the flow ID and the request ID. If you contact support about a failed run, include both: they let support trace the exact run.
|
How you ran it |
Where the error appears |
What happens to the work |
|---|---|---|
|
Machine translation (document preparation, the Machine translate action, or the Editor) |
The operation that started it reports the error: the Prepare files popup, the action's progress window, or the Editor. |
Nothing is written. Machine translation delivers its results in one piece at the end, so a failed run leaves your content untouched, even a run that was nearly finished. |
|
Post-editing, as a supplier service |
The error is recorded in the job messages, and the AI work status popup (opened from the sparkle icon) shows the failed run. The job stays In Progress: a job that does not complete is your signal to open its messages. |
Segments processed before the failure keep their post-edits. |
|
Quality assurance, from the Editor (Run QA) |
The QA run reports the failure in the QA panel. |
Findings recorded before the failure remain, and the run is reported as failed, never as a finished check. |
|
Quality assurance, as a supplier service |
Same as post-editing: job messages and the AI work status popup, and the job stays In Progress. |
Same as above: partial findings remain, and the run reads as failed. |
A workflow reports Completed only when the work genuinely completed, so a completed status is one you can trust.
Cancelling AI Work
Cancelling AI work stops it. The AI engine is told to stop the flow: it finishes the batch it is working on, starts nothing further, and any result that arrives afterwards is discarded. This applies to both Standard and Codyt projects.
|
Where you cancel |
What stops |
|---|---|
|
Cancelling or deleting a job assigned to an AI supplier |
The flow behind that job: post-editing or AI quality assurance |
|
Cancelling a machine translation pretranslation task |
The machine translation flow |
|
Cancel operation on the Prepare file(s) for translation popup |
The machine translation flow that runs while a file is prepared |
|
Cancel operation on a QA run in the Editor |
The AI quality assurance flow |
What happens to the work already done depends on the flow:
-
Machine translation and post-editing: segments already written before the flow stopped are kept, and nothing further is written.
-
Quality assurance: the run is reported as cancelled and incomplete, and its partial findings are not presented as a finished check.
A stopped flow shows as Cancelled in the job's AI agent status, with no error attached. That is different from Failed, and the task or QA run reaches its final state straight away rather than staying in progress.