The article, if you read it, says that the 90 minutes is compared to those who don't use AI, i.e. a different population. I think they're asking whether those same people who use AI and work longer also worked longer before AI