Xerox PARC is a great example - PARC invented everything, but very little of it was immediately commercially applicable - Good management, is what turns "people producing great things", to "people producing great things regular people can use"
Anyone with a successful research lab is a manager: they're spending a good chunk of their time herding grad students, writing grant proposals, hiring junior faculty, etc. This model works.
JG is the reason Google has all that ML effort. He joined google via the MetaWeb acquisition; when he'd finished his earn-out he wanted to leave but was asked to look around for something useful to do and consolidated a bunch of dispersed ML efforts. He also bought Deep Mind on behalf of Google.
I don't know why you would say he's "not a DL/ML guy per se" -- he's a programmer, not an MBA type. You consider MetaWeb and TellMe inadequate?
So you concede that there was “magic” being done at Google, but you want to argue that Giannandrea’s leadership of the search and AI teams had nothing to do with it?
Don't you think it's likely that JG will convince other top AI/ML talent to join Apple? That's usually a senior exec's main job, to recruit, retain and grow people who are building the products.