Find the first failed action in a cloud flow
Read the earliest relevant failure, in execution order.
Applies to: Power Automate cloud flows.
Revision 5 · 18 November 2026 · General first-line guidance. Wording varies by version; a symptom is not a confirmed diagnosis.

What does the message mean?
Some rows describe a visible symptom rather than an exact message. Wording marked “varies” may differ on your screen.
| What you see | What it may indicate | What to check |
|---|---|---|
| Failed / skipped action | A skipped action can be a result of an earlier failure, not the original fault. | Read the earliest relevant failure, in execution order. |
| Failed | An action returned an error. The detail identifies the cause. | Read the action message; do not resubmit before checking side effects. |
| Running / waiting | The run has not finished. This is not automatically a failure. | Check the current action and timing before resubmitting. |
| No message at all | Problems can happen without any message. A blank screen or a missing control is still useful evidence. | Record the visible behaviour, the time and who is affected. Do not guess a code. |

Step-by-step checks
Work in order and stop when it is resolved. If your screen differs, note what is different and ask IT or the owner.
Open the correct run
- Do this
- Select the run that matches the event time and business item.
- What you should see
- The run time and trigger match the incident.
- Then what?
- If this explains what you see, record it and continue only if the next check still applies.
Inspect action states
- Do this
- Expand the first relevant failed action, not only the last red one.
- What you should see
- An action name and error message are visible.
- Then what?
- Record the result without changing settings. Stop if work, data or security could be affected.
Read inputs and outputs safely
- Do this
- Look at the relevant fields without copying tokens, personal data or full payloads into tickets.
- What you should see
- You can describe a missing or unexpected value.
- Then what?
- Use only the approved action described here. If it does not match your screen, stop and ask IT.
Check what already happened
- Do this
- Look for actions that already sent mail, created records or changed data.
- What you should see
- You know whether re-running could duplicate an effect.
- Then what?
- If unresolved, contact IT or the owner with sanitised evidence. Do not repeat actions that could duplicate data.
When to stop and contact IT or the owner
- Work could be lost, or you cannot tell which version is current.
- The problem repeats, or a managed setting blocks you.
- A step would change production data, shared connections, permissions or security.
What to put in your support request
The app and version if visible, your device, the time, the exact message, who is affected, and the checks you tried with what you saw. Remove names, addresses and confidential content from screenshots. Never send passwords or MFA codes.
Microsoft reference
Microsoft documentation for this topic (opens a new tab). Some links are product hubs. Official pages may include advanced steps; ask IT before applying them on a managed device.
Related guides
Was this guide helpful? Leave feedback
Tell us what worked, what was unclear, or what is missing. Feedback is private and goes to The Workflow Lab for review. It is not shown publicly and is not an IT support request.
The Workflow Lab is independent of Microsoft and is not your employer’s service desk.