I want to tell you about the fastest expensive hour I've ever seen, because the lesson in it applies to every business, not just ones running ads.
The setup
One of my clients runs serious paid advertising. Meta is one of the platforms. One day, a mistake on the platform side, not the client's, not mine, sent spending sideways. By the time a human looked at a dashboard, $60,000 was gone. Under an hour.
Let that number sit. Sixty grand. Not over a bad month. Not over a slow week of underperforming ads. In less time than a lunch meeting.
Nobody was negligent. That's the scary part.
Here's the thing I need you to understand: nobody did anything wrong. The client wasn't asleep at the wheel. I wasn't. The account was checked regularly, the way every well-run account on earth is checked. A human looks in the morning, maybe midday, maybe again before dinner.
That rhythm has been "responsible management" for as long as digital advertising has existed. And it has a hole in it you could drive a truck through: everything between the checks.
A human cannot stare at a dashboard every minute, around the clock, forever. Not because they're lazy. Because they're human. They have other work, they eat lunch, they sleep. The gap between checks isn't a flaw in your people. It's a flaw in asking people to do a machine's job.
What I built
Within 24 hours of that mistake, I had a monitor running. It's not fancy and that's the point:
- It watches spend against budget constantly. Not when someone remembers. Constantly.
- When spending moves in a way it shouldn't, it alerts immediately. Phone-buzz immediately, not tomorrow-morning immediately.
- It never gets bored, never takes a day off, and never assumes someone else is probably watching.
It has watched every dollar since. The $60K hour has not happened again, and if the conditions for it ever line up, the alert fires in minutes instead of the damage surfacing after an hour.
Total glamour factor: zero. It's a smoke detector. Nobody brags about their smoke detector. But nobody who's had a kitchen fire thinks it's optional either.
The best automations are boring guardrails
When people picture AI and automation, they picture the shiny stuff. Robots writing content, chatbots charming customers. And sure, some of that earns its keep.
But the highest-value automation I've ever shipped doesn't create anything. It just watches. And that's true for most businesses, because of one asymmetry: work that doesn't happen is visible, but a number nobody's watching is invisible. You notice when the report is late. You don't notice the thing silently going wrong between checks, until the invoice arrives.
A guardrail flips that. It makes the invisible thing loud, the moment it matters.
What a guardrail looks like in a normal business
You don't need a $60K ad budget for this to apply. Every business has numbers that only get looked at when someone remembers:
- Invoices. Which ones crossed 30 days unpaid this week? A watcher that flags them beats hoping the bookkeeper's Tuesday routine catches everything.
- Inventory. The ingredient or part that ran out and stopped a job. A threshold alert costs almost nothing.
- Bookings. A day that's suspiciously empty because the booking widget quietly broke on Thursday. Who would notice, and when?
- The website. If your contact form died right now, how many days of leads would vanish before anyone knew?
- Reviews. A one-star review sitting unanswered for two weeks does damage a same-day reply would have contained.
None of these need AI, exactly. They need something tireless with one job: watch this number, and yell if it moves wrong.
The question to take with you
So here it is: what's the one number in your business that nobody is watching between check-ins?
Not the number on your daily report. The one that would only surface after it already cost you. Every business has at least one. Mine did too, until an expensive hour pointed at it.
If you know your answer immediately, that's the guardrail to build first. If you're not sure, that's a different kind of answer, and honestly the more common one. Finding that number is exactly what a diagnostic is for.
The full case study is on the site: the ad spend monitor. Built in 24 hours, watching ever since.
Storm photo: Kelly DeLay (flickr.com/photos/ravedelay)