IMO the first obvious benefit of Generative AI is the ability to generate lots of variants with way-beyond-good-enough quality and with lots of variety, and at practically zero cost.
This has tremendous implications for “opening the diamond” of Design Thinking-inspired design processes, because it enables very wide exploration and elimination of directions that don’t seem to match the target.
So, instead of working on 2-3 options, you can now generate hundreds with the same or lower effort, cluster them, organize them in any way, and only then invest the actual human effort of polishing the better ones.
It’s in my mind the same idea as going from high-fidelity trial-and-error engineering simulations with high cost per variant towards a model of generating thousands of designs using parametric studies and low-fidelity yet accurate-enough simulation models, then eliminating regions of the design space, and progressively refining your search beore you splurge on verification with high-fidelity simulations or validation with hardware testing.
This takes quite a bit of the “dark art” out of engineering design (or design in general), which understandably disturbs many experts…
Such is life, though.