Snake Eyes Comes Up Way Too Often
The 2FA Code That Never Showed Up
I hate email. I've hated it for years, and if you ask Donita she'll tell you I can go thirty minutes on why the whole medium needs to die. But this week email taught me something I wasn't expecting, and it started when a 2FA code never showed up.
I was trying to authenticate into a service and needed the code, so I went to my inbox and found nothing there. I had to go in and manually tell the system to fetch new messages before the code finally appeared. Weird, but I moved on.
The Pattern I Keep Running Into
The next morning around 10 AM, I looked at my inbox and the most recent email was from the day before. I get 30 to 50 messages a day between newsletters, bank alerts, deploy notifications, and the usual noise, so an empty inbox at 10 AM means something is broken.
But I was heads-down on three other projects, so I prompted the system, watched the emails populate, and went back to building. This kept happening over the course of the day and I kept forgetting about it because the build always felt more urgent than the thing that was broken.
Here's the pattern I keep running into. I notice something is off, I know I should stop and fix it, and I don't.
Why Email Is Infrastructure, Not Annoyance
So why do I care about email if I hate it so much? Because in my world, emails are signals. Every inbound message is a piece of data that my system needs to automate my business accurately. A bank alert triggers a financial workflow, a deploy notification confirms a release went live, and an auth link unblocks a login. When those signals stop flowing, the whole chain downstream goes quiet, and quiet is dangerous when you've built your business to run on automation.
The emails weren't missing. They were being routed to junk, every one of them, because my system made a snap classification with bad information. Like a filing clerk who shoves the whole stack in the wrong drawer and goes to lunch.
The Numbers Behind Ungoverned AI
Ungoverned development is rolling the dice. Snake eyes comes up way too often.
78% of knowledge workers use AI tools at work, and most of those tools make the same mistake my email system made: classify fast, commit permanently, never look back.
70 to 85 percent of AI projects fail, and it's almost never the technology. It's the part nobody budgets for, which is what happens when the system is wrong and confident at the same time. The system will never tell you about it because it doesn't know it's wrong.
Organizations with formal governance are twice as likely to succeed with AI. That's why we built the governance layer nobody wanted to talk about because governance creates the one thing most automated systems skip. The second look.
**## The Detection-to-Fix Ratio
Every email, misclassified. Zero alerts fired.**
The fix took 22 minutes. Detection took however long it takes me to get annoyed enough to stop building and start looking.
That ratio is the whole lesson.
I spend most of my time now building governance for automated decisions — the code is writing itself and nobody is watching, and that pattern shows up everywhere, not what they decide but whether they get to decide once and walk away without checking.
The part that bugs me is the human side. I noticed multiple times, I knew something was wrong, and I still didn't stop because the build felt more important than the maintenance.
How many of your systems made a confident call today that nobody checked?
If that question makes you uncomfortable, that's the right reaction. Schedule a conversation and we'll talk about what a second look actually looks like.
