Just because one isn’t good at a thing doesn’t preclude one from being a sufficiently passable judge of a thing.
To wit, the answer pre-AI was to hire an expert on that thing, and you would then critically assess their work product, despite being unable to build it yourself.
True, but if you hire a generalist and they are consistently under-performing specifically in the subject matter where you are an expert, it may behoove you to take the rest of their work with a grain of salt as well.
The key is to understand what someone is actually good at, rather than lump them into some amorphous "generalist" category. Along with (presumptively) broad experience, a generalist is just a specialist at various things which often feel obtuse or reductive to delineate — e.g. "I'm a specialist at rapidly narrowing vague failures into specific causes, assessing scalability trade-offs, understanding edge-cases at the intersection of two programming languages, and optimising cache invalidation."
Perhaps the best generalist skill when working in teams of specilists is "a reasonably accurate bullshit detector."