The Krugerisms™ · Field Intelligence Series
The Most Expensive Phrase in Our Industry
Normalization of Deviance — what it is, what it’s quietly doing to your infrastructure right now, and the only methodology that stops it.
Mark Kruger·
48 years in the data and voice industry·
Founder, JustAskKruger.ai
“It works for now. Don’t worry about it — we don’t have the timeline, hours or the budget.”
Your best engineers end up permanently assigned to the past.
I’ve been in the data and voice industry for 48 years. Server rooms, IDF closets, equipment bays — new builds, takeovers, emergency calls, post-mortem audits. These days I walk into AWS consoles, hypervisors, and cloud tenants the same way, and I find the exact same thing. That one sentence has cost this industry more money, more downtime, and more credibility than any product defect, vendor failure, or network outage ever has.
Notice it has two halves — and they usually come out of two different mouths. “It works for now” is a tech reporting a condition. “We don’t have the timeline, hours or the budget” is someone else closing the ticket. Neither believes they did anything wrong. But in the space between those two sentences, a temporary condition just became permanent.
That handoff is where every single one of these starts.
Challenger Wasn’t an Accident
Sociologist Diane Vaughan spent years studying the Challenger disaster. What she found wasn’t negligence — it was something more dangerous. She called it Normalization of Deviance: the gradual process by which workarounds become accepted as normal because they haven’t failed yet.
NASA engineers knew the O-rings were eroding. Each time the shuttle came back safely, the risk got quietly absorbed into the baseline — until the morning it didn’t.
Challenger wasn’t an accident. It was a predictable failure. And I see that exact pattern every single day in our industry — in networks nobody would describe as unsafe, right up until they are.
The Deviance Nobody Walks Past
In today’s environment it comes in two forms — and the second one is the more dangerous one, because nobody ever physically lays eyes on it. The physical kind you can trip over. The NAS on a cable spool on the IDF floor…
🔓
Continue Reading — It’s Free
The next five sections cover Device Normalization™, The Three-For-One™, Say It Out Loud™, and Good Housekeeping™ — and why AI makes this problem worse before it makes it better. Enter your name and email to unlock instantly.
No spam. No sales calls. Just the article — and early access to JustAskKruger.ai when it launches.
The Deviance Nobody Walks Past
In today’s environment it comes in two forms — and the second is more dangerous because nobody ever physically lays eyes on it.
The physical kind — deviance you can trip over
- The NAS sitting on a cable spool on the floor of the IDF
- The switch with a failed redundant power supply that’s been “on order” since 2022
- The IDF door propped open with a fire extinguisher because the badge reader is slow
- The rack that never got its ground bond, on a site that’s been live for six years
- The UPS running on batteries four years past their replacement date
- Cable draped across ceiling tile because nobody wanted to hang J-hooks that day
- The voicemail box still active on a user who left the company in 2019
The virtual kind — the deviance nobody ever walks past
- The SIP trunk misconfigured for three years but working
- The security group opened “just for the cutover” in 2023, still wide open today
- The admin-level IAM role left behind by a migration that finished two years ago
- The DNS record still pointing at a host that was decommissioned last spring
- The test tenant that quietly became production because it happened to work
- The expired-cert alert that auto-renews via a script only one person understood — and they left
- The infrastructure-as-code state file that no longer matches what’s actually running
Notice what the second list has that the first one doesn’t: no witnesses. The cloud didn’t eliminate normalization of deviance — it removed the last thing that used to catch it.
And here’s the part that should bother you: the timeline you protected, the hours you saved, the budget you held — all of it was real. But it was a loan. Every one of these got borrowed against a future outage, and that note comes due at full interest, on a holiday weekend, at 2am, with a customer on the phone.
The Three-For-One™
Here’s the Krugerism that governs all of this: whatever you put off today will jump up and bite you — and it will cost more timeline, more hours, and more money than doing it right in the first place ever would have.
It doesn’t come back once. It comes back three times.
- The original work is gone. The timeline, hours, and budget you spent doing it wrong bought you nothing but a delay.
- You pay again to correct it — except now it’s in production with a customer on it.
- You pay to clean up behind it: the outage, the escalation, the after-hours call, the credibility you spend six months earning back.
Three invoices for one job — and only one of them was ever on the quote. So “we don’t have the timeline, hours or the budget” is never actually true. You always had it. You just committed it to the expensive version without knowing it.
Trust No One. Assume Nothing. Verify Everything.™
The cure has two hands. This is the first — the technical one. I call it Device Normalization™.
You go back to baseline systematically, deliberately, and without mercy. Every endpoint documented. Every firmware version verified. Every configuration confirmed against standard. In the virtual estate: every instance owned by a name, every rule justified, every credential traced to a person who still works there.
Krugerism #1
Trust No One. Assume Nothing. Verify Everything.™
Trust no one, because the last engineer normalized something and didn’t write it down. Assume nothing, because the drawing on the wall is four years old. Verify everything, because the only condition you actually know is the one you put your own eyes on. Bottom up. Layer 1 first.
Say It Out Loud™
This is the second hand — the one nobody wants to talk about. And it’s the half that killed Challenger. Vaughan’s finding wasn’t that NASA lacked data. They had the data. What NASA lacked was an environment where raising a concern outweighed the schedule.
Every one of us has been in that room. The junior tech who sees the problem but doesn’t want to look stupid. The site lead who knows the cutover isn’t ready. The engineer who documented the warning, got overruled, and quietly decided never to bother again. Silence is how deviance normalizes. Every time.
Say It Out Loud™ means building a room where the newest person can tell the most senior person “that’s wrong” and pay no price for it. If you’re the senior person in the room, this is entirely on you.
Good Housekeeping™ — and Why This Matters Now
If it’s worth plugging in, it’s worth mounting. If it’s worth mounting, it’s worth labeling. If it’s worth labeling, it’s worth documenting. Full stop. Not next week. Now.
AI cannot fix what it cannot see cleanly. You cannot build an intelligent fabric on top of decades of accumulated drift. And AI makes the human half worse before it makes it better — when a machine hands you an answer with total confidence, the instinct to question it drops through the floor. That’s normalization of deviance with a faster clock speed.
AI doesn’t solve normalization of deviance. Methodology does. AI just makes the methodology scalable.
This article is the foundation of a 7-part LinkedIn series. Each post drops Tuesday and Thursday mornings. Follow Mark Kruger on LinkedIn for the series — bookmark this page for the complete picture.