Skip to content
← All insights

Communication ·

When Microsoft 365 goes down, crisis communication must still work

by Federica Grazia Bartolini
CEO & Founder · FGB Communications

FGB Communications editorial illustration: Microsoft 365, continuity and communication
FGB Communications editorial illustration.

In brief

Microsoft announced service restoration on 14 November after a disruption affecting email and other workplace services. The incident puts alternative channels, access to the crisis plan and communication responsibilities under scrutiny. Service access and completion of backlogs are distinct parts of recovery.

When Outlook and Teams stop working, the issue is more than the number of meetings cancelled. A business can lose track of who knows what, which decisions have been made and how to reach employees, customers and suppliers.

The Microsoft 365 disruption that began on 12 November brought this dependency into focus. Operational continuity is part of the story, alongside communication and trust: a crisis plan has to remain usable when familiar tools are unavailable.

What is established

On 14 November, Reuters reported Microsoft’s announcement that services had been restored across affected customers and regions. The disruption affected Outlook, Teams and other products, with consequences for businesses and public organisations. Reuters also reports that Microsoft said it was not aware of malicious activity and ruled out AI as the cause.

The incident was tracked as MO1550821. Microsoft’s Service Health portal is the official operational reference. A public status page and a detailed post-incident report serve different purposes; the fuller review remains important to understanding the sequence of events.

I do not use a total estimate of people affected. Outage reports are not a census of users. Nor would I describe this as “all of Microsoft being down”: services, locations and recovery times can differ.

The dependency extends to AI integrations

OpenAI’s status history recorded errors involving its Microsoft 365 connector on 13 November, subsequently resolved. Adding an AI assistant to a workflow does not remove its reliance on connected systems being available.

That does not establish that AI caused the disruption. The organisational question is which activities can stop together because they share identity systems, archives, email or external services. A workflow that appears distributed may still depend on a common point.

Communication when the usual channel is unavailable

A practical starting point is access to the crisis team’s contacts. Who can retrieve them without signing into the affected environment? A plan stored only inside an unavailable system may be inaccessible precisely when it is needed.

An alternative channel has to be selected and approved beforehand. Its usefulness depends on whether the intended people can access it and whether it is suitable for the information being exchanged. Moving confidential documents to personal accounts can create another problem while trying to solve the first.

The immediate message is often simple: which activities are affected, what people can do, what to avoid and when the next update will arrive. A credible update cadence has more value than generic reassurance or a recovery time the organisation cannot guarantee.

Restoration does not automatically close the crisis

Regaining access does not establish that every message has arrived or every task has been completed. Backlogs, deadlines, customer requests and suspended decisions remain part of the recovery. Coordination across IT, operational teams, HR and communication matters here too.

My conclusion is that dependencies have to be understood, accepted and tested under realistic conditions. Adding tools without examining their independence can increase complexity without improving resilience.

A short exercise could begin with one question: if email, chat and the shared archive were unavailable tomorrow, which decision could we still make within an hour? The answers reveal more than the existence of a document called a crisis plan.

Sources and update

Checked 14 November 2026: Reuters on service restoration, Microsoft Service Health, OpenAI connector incident history. Outage reports are not treated as a count of affected users. Organisational considerations are the author’s analysis.