You agree AI improves engineer productivity. So last remaining question is, does greater productivity mean that fewer people are required to satisfy a given demand?

The answer is yes of course. So at this point, supporting the assertion requires handwaving about shortages and induced demand and demand for engineers to develop and support AI and so on. Which are all reasonable, but it should become pretty apparent that you can't be confident in an assertion like that. I would say it's pretty likely that AI has resulted in engineers being laid off in specific instances if not the net numbers.