The Short Answer
Check What Already Happened Before You Retry
If a business automation fails, check three things before running it again: did the workflow start, where did it stop, and what has already happened in the other apps? Find the original enquiry, order or job, then compare the automation's run history with the records in the apps it updates. A run can fail after it has already created a job or sent an email.
In this fictional example, a Sunshine Coast maintenance business turns accepted quotes into scheduled jobs automatically. The office can see a failed run, but the job already exists. It was the confirmation email that failed. Starting the whole workflow again could put the same customer on the schedule twice.
What Matters Most
- Trace one affected reference from the trigger to the destination.
- Mark each action done, missing or unknown.
- Reconcile the manual queue before restarting the workflow.
Business AI Automation · Your Next Step
Start With One Repeated Task, Not An AI Shopping List
Describe the task, its frequency and the tools involved. We can compare a reliable rules-based workflow with AI assistance and agree how the result would be checked.
No promised savings before measurement. Access, sensitive data, human review and ongoing platform costs need an agreed scope.
Enquire first. Scope and costs are agreed before work begins.
Explore The Service And Inclusions01
1. Find The First Missing Step
Pick one real example and note its reference, the time and time zone, and what you expected to see. Use that to follow the job through the apps. The trigger is the event that starts the automation, such as a quote changing to Accepted.
| What You Can See | What To Check Next |
|---|---|
| The quote is accepted, but there is no run | Check the workflow is enabled, the trigger is watching the right account or list, and the record met its rules. |
| A run stopped before the job was created | Look at the first failed action. Check for missing fields, an expired connection or a temporary service error. Confirm the destination has no matching job before replaying. |
| The run stopped after a job was created | Keep the existing job. Identify the missing step and complete it once. |
| The run shows Success, but the office has no job | Search the destination by quote reference. Check whether a rule skipped the job step, or the workflow wrote to the wrong account or list. |

02
2. Recover The Work Without Creating Duplicates
Write down the outcome of each action: done, missing or unknown. Check the actual job system, sent mail or invoice records. A timeout means the automation did not get an answer in time; the other app may still have accepted the request.
If you can confirm the job exists and the customer has received no confirmation, send the missing confirmation through your approved process. If you cannot tell whether a payment, invoice or message went through, check the destination or ask the responsible person to confirm before repeating it.
If several records are affected, agree who will handle them and where they will track them. If the workflow is actively creating duplicates, have its owner pause the affected automation and put a manual queue in place for new work. Record what was handled manually so it does not get processed again when the workflow comes back.
For the quote example, a recovery note could read: quote Q1042, job J271 confirmed, confirmation missing, office lead to send it once and mark it done. Before switching the automation back on, check how it handles work received during the pause. Compare that backlog with the manual queue and confirm that each accepted quote has one job.
03
3. Make The Next Failure Easier To Spot
Give each enquiry or job a reference that follows it across your tools. Where the tools support it, have the workflow use that reference to find an existing record before it creates a new one. Ask your developer how it behaves if two attempts arrive at the same time.
Set up an alert that goes to someone who can act on it. Include the reference, the step that failed and where to check it. Keep customer details out of alerts unless the recipient needs them and the channel is approved. Also decide how to catch work that never started: for example, compare accepted quotes with created jobs at the end of the day.
Test recovery with a dummy record: one normal run, one missing field and one failure after a record has already been created. Check that staff can find the failed item, finish it once and tell when the queue is clear. Microsoft's error handling guidance covers retries and failure notifications for Power Automate.
04
Want Help Making An Automation Easier To Run?
Ultra IT's developers on the Sunshine Coast can help trace where the work stops and scope changes such as clearer alerts, a review queue or a safer way to restart a job. Ask Ultra IT to review an automation. Share the apps involved and one example with sensitive details removed. If you're still choosing a process to automate, start with our guide to which business processes to automate first.
Common Questions
Clear Answers Before You Decide
What If The AI Step Gives The Wrong Answer?
Treat it as a workflow problem even if the run shows Success. If the AI put an enquiry in the wrong category, or drafted a message with the wrong details, give a person a place to review and correct it. Where the consequences are significant, have someone approve the result before it reaches a customer or changes a record.
Can I Turn On Automatic Retries?
Retries can help with temporary connection errors when the action is safe to repeat. They generally cannot fix missing information or expired access. For anything that sends a message, issues an invoice or creates a job, check how the destination handles repeated requests before relying on retries.
What Should I Save Before Changing The Workflow?
Keep the run reference, the first error, the version that was running and the affected job references. Store them somewhere your team is authorised to access. Do not export an entire customer record if a reference and a redacted error are enough.
Sources And Review Notes
This guide combines Ultra IT's practical service experience with the current primary sources below. Product availability, pricing and platform guidance can change; check the linked source or ask us to review the current environment before acting.





