You know what they say about the first 90% taking up 90% of the development time and the last 10% taking up 90% of the development time. That is likely what we are seeing with current AI efforts. They get amazingly close yet aren't good enough to be relied upon.
I'm guessing that you didn't mean to end up with 180% of development time, but that does seem about right when it comes to the software development projects that I've been part of. I also think you're correct that despite the impressive recent advances in AI, the remaining effort may still be enormous.