No, foundation models are not the world-changing machines their evangelists promise. They are extraordinary plumbing, and plumbing changes daily life without granting its owners a seat in the room where decisions get made.

The pattern I keep seeing is this: companies buy access, paste prompts into a chat window, and call the result “AI transformation.” Meanwhile, nobody has rewritten the workflow, retrained the staff, or removed a single meeting. That is not transformation; it’s a very polished autocomplete with an API bill.

Yes, foundation models changed research, translation, coding, and creative work. Great. A model that drafts boilerplate or explains an error saves real time. But convenience is not revolution, and “it generates text” is not a business model.

The bigger problem is that we keep asking models to imitate expertise instead of giving people better judgment. The best deployments narrow the task, expose uncertainty, and leave accountability human. The worst ones produce confident nonsense and call it “scale.”

Foundation models are a genuine technological leap, not automatic social one. Use them, but stop dressing efficiency up as destiny. The future won’t be transformed by prompts; it’ll be transformed by organizations brave enough to change how work actually gets done.