Wedlake Consulting
Menu
Book a free intro →

Wedlake Consulting · Remote or in Manchester

← Notes

After a Microsoft 365 outage, check what your automations already did

When Outlook returns, some orders and approvals may still be waiting. How to keep urgent work moving and catch up without processing it twice.

Published
Reading
4 min
Topic
Automation

On 12 November 2026, a Microsoft 365 outage disrupted access to Exchange Online mailboxes. In its service health updates, Microsoft also reported that email delivery could be delayed, requests for mailbox data could time out and some Teams calendar and scheduling features could fail. Microsoft service health updates, archived by the University of Pennsylvania

Think of a shared inbox where a Power Automate flow picks up incoming orders, creates records in your order system and sends acknowledgements. During an outage, one order might still be waiting in the mail queue. Another could already be in your order system, with only the acknowledgement outstanding. Both customers might call to ask whether you received their order.

I would start with those orders. Give the team a way to record urgent work while email is unreliable, then work out what happened to each order before you run anything again.

Keep urgent work moving

Ask your IT provider to check Health → Service health in the Microsoft 365 admin centre and note the incident reference and estimated start time. In parallel, ask your operations team which jobs depend on the affected services. Start with jobs that have a deadline or cause a customer problem if nobody acts. A daily summary can wait while the team checks orders due for dispatch. Microsoft’s guide to checking service health

Agree one fallback route that works without the affected service. That could be an existing business phone line for urgent orders, with a nominated colleague recording them in the order system if it is available. If you need a temporary log, use an approved place the team can reach during the outage. For each order, record its reference, when it arrived, who took it and what they did. Give the colleague taking calls a way to check whether an order already exists before they create it.

Keep your normal approval rules in place. If an invoice needs authorisation and the approved fallback is unavailable, record the deadline and hold it for the approver. Make sure whoever will clear the backlog can see those decisions.

Match the work against what actually happened

Once the affected services are available to your team, compare the original requests, the fallback log and the records in the system where the work lands. Start with the estimated incident window, then include messages that arrive late. The mailbox and run history are both useful here. An incoming order that never started a flow can be missing from the run history entirely.

For runs you can find, ask the person who looks after the flow to open the relevant steps. Microsoft documents how to inspect a failed step and resubmit a run. I would put a check of the destination system between those two actions. Microsoft’s guide to troubleshooting a cloud flow

  1. Inspect the run historyFind the order and check how far the flow got.
  2. Check the destinationLook for the order record and any action taken by hand.
  3. Complete the outstanding workResubmit only if the earlier steps are safe to repeat.
The order I would use to check work before resubmitting a Power Automate run. Source: Microsoft's cloud flow troubleshooting guide.

If the order exists and the acknowledgement is the only thing outstanding, check whether it is already queued before sending one. If an action timed out, check whether the destination accepted the change before the reply was lost. If the order was entered by phone and the original email later turns up, match it to the order already on file. Put uncertain cases to one side for someone to check with the order system owner before releasing them.

Test one low-risk item end to end before clearing the rest. Check the business result as well as the run status. For an order, that means the correct record exists once and any required acknowledgement is accounted for. Work through the remaining items in small batches and record how each was resolved.

Make the next outage easier to recover from

Pick one automation during November and ask whoever built it to show you what happens if the mailbox is unavailable or the final step fails after creating a record. Use test data and ask someone from operations to take the job through the fallback and back into the normal process.

I would give each request a reference that follows it through both routes. Have the automation check that reference against existing records, and ask the builder to show how the order system prevents duplicate creation if two attempts arrive together. If the system cannot do that, agree who reviews repeat attempts before the work proceeds.

Write the fallback and the recovery checks where staff can reach them during a Microsoft 365 outage. When you test the process, hand a colleague an order that was entered by phone and has since arrived by email. Ask them to show how they would get it to dispatch once.

Written by

Owain Wedlake, technology and AI consultant in Manchester. Questions or a different view? Reply by email.

Next step

Start with a conversation.

A free intro session. Bring one process that eats your team's time, and leave with a straight answer on what's worth changing.

Free · No obligation · Remote or in Manchester