Nobody Noticed the AI Agents. That’s the Part That Should Scare You.
AI agents read confidential records and no one noticed. The problem wasn’t the AI. It was the credentials left lying around.
[Click on the image to view post]
Educational Content
Your AI Isn’t Forgetting. It Ran Out of Room.
Someone asked me what an AI token is and how it’s used. Here’s the plain answer, and why the context window shapes every AI answer you get.
[Click on the image to view post]
The AI Was the Cheap Part.
Before you approve the AI budget, check what it leaves out. The real cost of AI is everything you build around the model. Here is how to find it.
[Click on the image to view…
What Work Should an AI Agent Do? A Five-Question Test
Before you buy an AI agent, score the work: can you undo it, how bad can it get, and who owns the fix? A five-question test for business leaders.
[Click on the image to view…
Who’s On Call for Your AI Agents?
Ask who gets paged when a server fails at 2 a.m. and you’ll get a name instantly. Ask who gets paged when an AI agent misbehaves and watch the pause. IBM expects most large enterprises…
25 Clients. The Same Three Things Break Every Time.
A CEO asked what I expected to find in his assessment before we
started. I named three things. He pushed back on all three. Six weeks later, all
three were true. Across 25 active clients…
Google Stopped Sending People. Your Website Was Built for a Model That’s Ending.
Traffic down 33–38% and nothing broke. Why AI search traffic ended the referral deal your website was built on, and what to instrument before you panic.
[Click on the image to view post]
Nobody Offboards an AI Agent
When a contractor rolls off your team, a clock starts — badge, VPN, repos, mailbox. Somebody signs a form. Now answer this: when you give an AI agent access to your systems, what’s your process…
We Didn’t Fix More Incidents. We Just Saw Them Sooner.
The client asked for a 25% improvement in incident detection. We delivered 38%, and more than $2 million in documented savings followed. What surprises people is that we never got meaningfully better at fixing things…
Your AI Guardrails Are Written in English. That’s the Problem.
In July, OpenAI and Anthropic each disclosed that their AI models escaped supposedly sealed test environments and broke into real companies’ production systems. The models weren’t hostile. The sandbox was misconfigured, and the guardrail was…