This story comes from someone we know and stay in close contact with, rather than from our own office. We are telling it as he and his team experienced that Friday, because the scene was much the same in many offices that day.
Friday, just after eight: the inbox keeps spinning
Friday morning at the office: coffee, computer on, open Outlook. Then came the message “Loading profile …”. And there it stayed.
Anyone who has worked in an office for a while knows what happens next. Close Outlook and open it again. Check the Wi-Fi. Look over at the next desk: “Is yours working?” Nobody’s was. By the time the first computer was restarting for the third time, it was clear that the problem was outside our office.
Microsoft’s status page confirmed it. Exchange Online, the service behind Outlook, had been experiencing a worldwide outage since Thursday evening. We simply had not noticed because we had already finished work by then.
What worked and what did not
And LIVOI? The platform was running, the models were responding, and everything already stored in LIVOI could be searched and analysed as usual. Only new information from Outlook was failing to arrive. Microsoft had closed the door to the inbox, and no tool can collect anything through a closed door. We were in good company: Microsoft’s own Copilot and the Microsoft integrations in Claude and ChatGPT could not access inboxes or calendars during those hours either.
Some people online celebrated the outage as an unexpected day off. That was not everyone’s experience. In businesses where invoices, orders and customer enquiries arrive by email, Friday mainly meant making phone calls, asking people to wait and catching up later. Accounting, assistants and office managers in particular had more work that day, rather than less.
What actually happened at Microsoft
The outage began on Thursday at 5:17 p.m. German time. Users could no longer access their Exchange inboxes, whether through Outlook in a browser, on a computer or on a smartphone. At its peak, Downdetector received around 130,000 reports for Outlook alone. Calendar functions in Teams were also affected, along with Copilot, SharePoint, Planner and several security services.
Microsoft identified an overloaded server process responsible for establishing secure connections as the cause. The start of the disruption coincided with a network configuration change, which Microsoft rolled back. A particularly persistent problem was that repaired servers failed again as soon as they came under load. Microsoft ruled out an attack.
On Friday morning, Microsoft reported the first stable results. Consumer services were working again, while business Microsoft 365 services were still restricted at that point. Microsoft announced that it would publish a detailed report on the cause.
One thing matters to us here: this was not a failure that an individual business could have prevented. Neither we nor our customers had any influence over it. At times like these, many specialists at Microsoft are working on a solution. Anyone who has ever restarted a server and hoped for the best knows how that feels.
What we take away from that Friday
After a morning like that, the question is not which provider is better. A more useful question is: what stops working in our business when a single service goes down? Three things became clear to us.
Know what depends on your inbox. In many businesses, incoming invoices, quotations, appointment arrangements and complaints all pass through the same inbox. People usually only notice when it is unavailable. A simple list is enough to show what will be left waiting on a day like this.
Have another way to get in touch. Businesses that receive orders and invoices by email every day should decide in advance how to reach their team and key customers when email and calendars fail. A phone list and a channel outside Microsoft 365 are often enough.
Ask what your software depends on. This applies to every solution, including AI assistants. We built LIVOI so that it is not tied to any single model provider and can connect to multiple sources, such as Outlook, Gmail, DATEV, OneDrive, SharePoint, Teams or WhatsApp. Information from Outlook will naturally have to wait for Outlook during an outage. But you decide whether everything has to pass through a single channel.
This is exactly the question we include in every potential analysis. We look at where AI can actually help a business, as well as what a solution depends on and what happens when one component becomes unavailable. You do not need any prior AI knowledge.
Conclusion: a quiet Friday with a lesson
In the end, it was a morning when nothing broke on our side, apart from an inbox outside our control. A few emails arrived later; everything else kept running. The real insight came afterwards: it takes an outage to realise how much depends on a single service.
If that Friday has left you wondering what your workflows depend on, or if you want to introduce AI in the office without creating new dependencies, get in touch. In a potential analysis, we work through what is worthwhile and what needs to keep running even on a day like this.