Maybe it's less about "what they can muster" and more about "what they think is sufficient to appease their investors who are vocally hype-chasing without diverting too many resources"? I think there's merit in sticking with "boring" applications that you have more confidence in there being potential utility in than trying to turn every problem you have into AI-hammerable nails.
One of the more useful things I do manually asking these lines is post long zoom transcripts into Claude and have it summarize it and answer follow up questions. I wish Zoom had it built in.