I've used it in my work and it feels like I haven't done any work. Other coworkers ask me for prompt engineering techniques because I happen to read books about this all the time. I have it do deep research, a paid feature, and it is surprisingly accurate. It is now accurate about 80 percent of the time which is a big improvement form about 60 percent in the inception of ChatGPT. It hallucinates a lot less. I test it's answers based on duckduckgo searches and I usually scroll down long enough to find the confirmation. AI companies sometimes hire experts to rate answers. Although AI is still in its infancy, it seems to do basic tasks with ease, medium tasks with least trouble and the paid version you can upload a PDF or a csv file and ask it about it. It rarely fails to answer basic lookup queries with these files. And this definitely saves a lot of time. There is a right and a wrong way to use AI. But I definitely can see where you are coming from.