For decades, organisations have been atomising the software engineering role into ever-narrower specialisations. Product owners took requirements, architects took design, quality engineers took testing, platform engineers took operations. What remained for many software engineers was writing code - and with generative AI, that's now the most automatable task of all. AI adoption among engineers is high. But the expected returns aren't materialising evenly, and measuring activity isn't telling us why. My Masters research, surveying professional software engineers across 28 countries, surfaced an unexpected finding about what actually predicts how engineers experience this shift - and it isn't something that shows up on a dashboard.

This talk connects that finding with twenty years of lived experience across startups and large enterprises to make a case that the way we've defined the software engineering role may be shaping the returns we're getting from AI. The good news is that systems can be redesigned.