“Some services aren’t available. We’re working on them. Please try back later.”
That’s the message some Chase customers got on Monday when they tried to log in. Bank of America, Wells Fargo, U.S. Bank and Zelle users were reporting problems at the same time. By Tuesday morning things were back to normal.
It was the third big outage in last five days. On Thursday, Outlook and Teams went down for Microsoft 365 users around the world, and it took until Saturday morning for everything to fully recover. On Friday afternoon, the FAA grounded every departing flight in the US for more than four hours.
Two things stood out to me.
First, nobody hacked anything. Microsoft, the FAA and Chase all said so. What happened is more ordinary. The FAA’s contractor was replacing old copper lines with fiber at a remote site, and a router there started overloading the air traffic control network. Microsoft said its outage came from an earlier network capacity upgrade. Chase pointed to a software bug in its own hardware.
So two of the three started with someone upgrading something. Anyone who has run production systems knows this pattern. Google even put a number on it in its SRE book: roughly 70% of outages come from changes to a live system. Big systems change all the time, and every change is a chance for something to go wrong.
Second, there’s nothing you or I can do to stop the next one. I can’t stop Microsoft from pushing a network change, and you can’t tell the FAA to keep its copper wires. The only part we control is how fast we find out.
The big outages make the news, so you hear about them eventually. But plenty of outages that cost a business money never make the news. Your payment processor has a bad hour, or your own checkout page breaks after an update. You find out from a customer, or later from a dip in sales.
How I’d set this up in ORION
Yes, I’m biased here. We built ORION to watch websites for you. You paste a page and tell it in plain words what matters, and it alerts you with an explanation of what changed.
For outages, I’d start with:
Status pages for the services you can’t run without. Many companies have one, often at status.companyname.com.
Your own homepage, login and checkout pages, and any API you use.
Vendor incident feeds, if they publish RSS or JSON.
If you use Claude or another AI assistant, you can connect ORION through MCP and ask it in plain words to add a Watch or tell you what changed this week.
If you want to try it today, go to orion.hexowatch.com, write down the five services your business can’t go a day without, and add their status pages first.


