The post expands on the thread here, for those interested in the context.
AI productivity tools seem like they work when deployed in the squishy ambiguous realm of "strategy" so managers don't realize that during execution, these tools not only fail, but you can *immediately* tell that they failed.
Of course, managers who try to replace their team with AI quickly find that it is completely incapable of passing the same test, for three reasons.
One is that AI can *do tasks* but expertise is not just a string of tasks.
Two is hallucination; the tasks it does do, it doesn't do well.