Member-only story
We completely done with LLMs in the enterprise
The gap between the boardroom narrative and the production reality has become impossible to ignore. The numbers are in. The experiments are over. And most of them failed.
There is a meeting happening right now at a company near you. Someone with a title that includes the word “transformation” is explaining to a room of tired people why the AI project they approved eighteen months ago hasn’t delivered what the slide deck promised. The slides were beautiful. The vendor demo was impressive. The pilot results looked encouraging enough to approve a budget. And then it hit production. And almost nothing worked the way it was supposed to.
I’ve had that meeting. I’ve sat on both sides of that table. And I’m done pretending that the enterprise LLM experiment is going the way anyone hoped it would.
This isn’t a column about AI doom, or about blaming the technology. The models themselves are genuinely impressive. What I’m talking about is the specific, structured mismatch between what large language models are built to do and what enterprise software actually requires. That mismatch is now producing real costs, real failures, and real decisions by real companies to quietly shelve initiatives they announced with enormous fanfare.
