The impression I'm getting over this huge number of AI roll backs is that AI is useful in some circumstances, but it's just not a cure-all. It is expensive, and increasingly so, straining the ROI scenario. My expectation is that the use-case for successful AI implementations is ultimately going to be narrow.