Untitled

"A $50-billion industry has spent decades trying to kill the spreadsheet. It's still here — and that's not a failure of adoption. It's the physics showing through."

"The names of the technologies change. The variable that decides the outcome does not."

"Different industries, different experts, opposite starting points — and the same conclusion. After nearly three decades of documented outcomes, the convergence itself is the signal."

One of the leading AI companies just acknowledged, on its own support page, that a single model can be confidently wrong — and told users not to rely on it as a single source of truth. That isn't a reason to trust AI less. It's the exact limitation I've been building around since 1998.

How HP proved technology is necessary and never sufficient.

"For Want of a Nail: Toyota's Fall and Climb to the 2026 Gartner Pyramid."

"Understanding the ubiquitous dot on Gartner's Hype Cycle — a graphic unchanged since 1995 — and what it means for AI success in 2026."

"A leading analyst just described the org structure of the AI era. I drew it in 1999 — before the cloud, before SaaS, before any of it existed. Here's the one condition his model still needs."

"Braum's bounded itself to a 300-mile circle, and the market calls it caution. It's the most disciplined growth decision in American retail — and the same law that governed a cotton exchange in 1931."

"I've had this conversation with CIOs for forty-three years. Last week I had it with an AI — and it made the same mistake."

"Strip the branding off a dozen AI maturity models and the skeleton is identical. Here's the one layer none of them measures."

"We've governed suppliers, couriers, and employees for decades. Then an AI agent enters the room — and somehow all the questions vanish. That's the entire mistake."

"'The AI said so' is not something you can defend to a board, a regulator, or an audit committee. This session is about what you say instead."

Here is the updated version with the corresponding graphic - by the way, this is for the Procurement Insights blog.

We didn't predict agentic AI. We built the model that agentic AI now has to live inside — twenty-seven years ago.

In January 2025 I asked whether any AI could tell me what time of day orders come in. One did — right after I told it order timing was the variable. That one-second gap, between answering a question and knowing which question to ask, is why adding an AI agent to an undiagnosed field doesn't build an intelligent ecosystem. It builds a faster version of the existing misalignment. Gemba walks one floor. The agent field has no floor.

Why IT keeps getting blamed for failures that were decided before the platform was ever chosen.