But it will do it adequately, expediently, and repeatably.

LLMs have a key weakness in that they aren't repetitive. They come up with a new solution to the job every single time they're prompted. Sometimes it can do the job adequately and expediently, sometimes it produces nonsense, and you'll never know when.

The c-suite have convinced themselves that LLMs don't have this flaw and that they can be treated like any other automation, that they can be relied on to do the same thing over and over, and then reality intrudes when they find out these are probability engines.

The chatbots can't actually do any job.