What Kept Running During the Microsoft Outage, and What We're Changing

This is Part 2 of our series on the November 2026 Microsoft 365 outage. Part 1 covered why every business needs a second way for customers to reach it. This time we're turning the question on ourselves: what happened to Zandolie's own systems while Microsoft was down?
What kept running
The AI agent didn't depend on Microsoft. Zandolie's AI agent answers on WhatsApp, through Meta's official WhatsApp Business Platform, and runs on our own servers. Nothing in that chain runs on Microsoft, so the outage had no way to stop it answering customers about stock, prices, opening hours and bookings.
Our own email kept working too. Zandolie's mailboxes don't run on Microsoft, so our inbox stayed up. But that turned out to matter less than you'd think.
Your email can be fine and still fail you
Email always has two ends. Even with our mail working, any message we sent to someone on Microsoft 365 was stuck on their side until Microsoft recovered. Microsoft held delayed mail for up to 48 hours, so some of it arrived the next day.
That's the part most businesses miss. Moving your own email to a different provider doesn't protect you if your customers, suppliers or staff are on the one that's down.
The weak spot: staff alerts
When a customer needs a person (a final price, a complaint, a test drive), the Zandolie AI agent hands the conversation to the team and sends a staff alert. Today, those alerts go by email.
If a client's team is on Outlook, an outage like this one means the AI agent knows a customer is waiting, but the people who need to know may not find out for hours. The conversation is safe in the Zandolie inbox, and staff who open the inbox see it straight away. But a handoff that relies on someone happening to check is not good enough.
Did any of our alerts reach a client's Outlook late during the outage? Honestly, we can't say for certain. Our records show when our email provider accepted an alert, not when Microsoft delivered it to the mailbox. That gap is the real lesson: an alert you can't confirm was delivered isn't good enough for a customer who's waiting.
Our email provider's records show none of our alerts bounced during the outage. That's reassuring, but it only tells us Microsoft accepted them, not when they reached anyone's inbox.
What we're changing
- A second alert channel. Staff alerts will go to more than one place, so a handoff never depends on a single provider. WhatsApp alerts to staff and Slack alerts are next on our roadmap.
- Visible waiting time. The inbox already shows when a customer's last message hasn't had a reply. We're making that harder to miss during long handoffs.
- An outage check in our runbook. When a major provider goes down, we'll check whether our clients' staff use it, and tell them where to watch for handoffs until it's back.
The lesson for any business
Ask the same questions we did:
- What keeps working if your email provider goes down?
- Who needs to be told when a customer is waiting, and how do they find out?
- Does that alert depend on the same provider as everything else?
If every answer leads back to one company, that's your single point of failure.
Coming next: Part 3
Microsoft published its preliminary incident report on 15 November, and its final report is due by 20 November. When the final report is out, Part 3 will cover what actually caused the outage, why recovery took so long, and what any business can borrow for its own incident plan.
See what Zandolie can do for your customers.
Get started