It’s so easy to get AI to create crap in volume. I can usually spot it. What I’m thinking about more is when I dont spot it, and believe it to be authentic human writing. That seems like the goal.
At this point it takes - for me - almost as much time to gen and edit, than to just author. Which leads me to wonder - if I heavily edit to the point of replacing it all, what was the point.
A good explanation I’ve heard is the “blank page problem” and that’s real. I can see the utility of using it to get started. But if you don’t heavily edit? Is AI generated junk really useful to your product team? I’m skeptical.
I saw a PM recently, under pressure to produce, use it to generate user stories in bulk. Reams of paper. 100s of pages. Which then got handed over to the dev team - who reacted accordingly. (Suspend disbelief on whether this is a good way to run a team - it’s not). Zero edits. 100% junk and it set them back weeks.
I’ve seen job descriptions written the same way which are wordy junk. When editing these it is basically a do-over. Making me wonder what the expectation was to begin with. Hiring is the single most important decision you make - why would you let a generic AI do it for you? Makes no sense.
I think the category error that a lot of software teams make right now is that they view AI as having “the answer” rather than thinking of AI as a sparring partner - a way to sharpen thinking.
>> JG - 3/2/25 - https://www.bellemonti.com