Most companies are finally bumping into something they should have figured out a while ago: slapping AI onto a broken process does not fix the process, it just makes the broken thing run faster. For the past two years, the default move has been to add copilots, assistants, and automation layers on top of workflows that were already held together with duct tape and institutional memory. Productivity metrics looked okay, but nothing actually changed.
The article draws a sharp parallel to the business process reengineering wave of the 1990s, and it is a fair comparison. Back then, the pitch was to redesign companies around information systems rather than bolt technology onto existing structures. It sounded radical. In practice, most of it became expensive reorganization theater because the underlying systems were too rigid to actually adapt. Sound familiar?
What is shifting now is the question itself. Companies that are getting somewhere with AI are not asking how to use it inside their current setup; they are asking whether the current setup needs to exist at all. That is a much harder question, and it has real consequences for anyone whose job is to design, run, or advise on how work gets done. The people who understand this distinction are about to become very valuable, and the ones still pitching AI as a productivity layer are going to spend a lot of time explaining why nothing moved.