The pitch for artificial intelligence has been consistent since the first large language models started capturing public attention: this is a tool, not a replacement. It will make you more productive. The people who use it will outcompete the people who don't. Everyone moves up, and the economy grows to accommodate the change, just as it did with every previous wave of automation.
The history of automation is where this argument gets complicated. It is true that every previous major technological displacement ultimately produced more jobs than it eliminated, in aggregate, over the long run. It is also true that the workers displaced by those transitions did not automatically migrate into the new sectors that emerged. The loom operators did not naturally become software engineers. The assembly line workers did not easily transition into financial services. What actually happened was decades of dislocation, concentrated in specific communities and specific demographics, with the costs borne by people who had no political or economic power to distribute them more widely.
Artificial intelligence is different from prior automation in one significant way: it goes after white-collar cognitive work for the first time. Every other automation wave hit manufacturing, agriculture, or low-wage service work first. This one hits legal research, medical coding, financial analysis, customer service, content production, and software development simultaneously. The people in those fields have more political voice than factory workers did. What they do with that voice remains to be seen.
What the actual labor market data shows is different from what the marketing says. We got into both, and into what serious policy responses would look like if anyone with the power to design them can stop arguing about whether the problem is real long enough to try.