If DEI fails to increase minority representation at a certain job without compromising on quality of candidate, do you believe it was unsuccessful?
That's overly simplistic. As johnny karate has pointed out, the purpose of DEI is to increase the pool of candidates. In my view, that further includes removal of all artificial barriers to academic, sports and general intellectual achievement that can be ascribed to policy error, bias, or correctable structural causes. All three of those causes are, in principle, realistically manageable, or can be compensated for (quality control, law). In short, make sure all your Einsteins get a reasonable opportunity to develop, then to be potentially chosen on merit, thus maximizing societal returns. Furthemore, those barriers that result from systemic bias are appalling and an affront to democratic principle.
Addendum
Maximizing the quality of the candidate pool should be considered part of what I refer to as full-spectrum capitalism, in that returns on not only physical and monetary capital, but also intellectual capital (human, artificial and IP stock), customer capital (relationships), and social capital (company and brand reputation; see Musk-Tesla), if managed properly, maximize returns on financial capital in the mid and long term. From a societal point of view, this is why education has traditionally been seen as key, its quality maximized.
Indeed, any company today selling products or services that can be subject to development or to replacement by alternatives needs to understand the real business they are in is in managing their intellectual capital stock, its inputs, and its quality control. From that, products and services can be seen as use cases of the output from core company competencies, targeted at or tailored to the users in those cases. (Somewhat fuzzier in consumer goods, as FMCG and entertainment are oft driven by the fadish and flimsy, built on ideas that go poof later. Still brand capital, tho.)
Unless you think AI will reach AGI (I do not) and that robots can do it all (I agree in principle), managing intellectual capital is key. If you want rich LLMs to make your AI more intelligent, you'll still need to feed it human intelligence. The rapid 0-to-100 rise of AI is perceived as wondrous, yet how will growth occur without new data? Tuning/feeding AI systems and their quality control appear to be tasks that will be around for a while.