Concrete examples are difficult in the same way that concrete examples for the abuses of "non-democratic societies" are difficult. I mean, it's not necessarily difficult to find some, but it's difficult to see the big picture from most examples.
Yes, FOX news already happens, but I would argue it's not a good thing, and making it more effective thus probably is even worse?
And yes, I guess much of the risk in a way are "tacky, classless acts by greedy shortsighted people", but that does not mean that they don't have any real consequences. Corruption is similar - and the effects in some economies are quite devastating, even though the individual bribes might not be that expensive. Big effects can arise from small individual inconveniencees.
But I also think that much of the risk lies in the future, with improved analysis algorithms. I think a reasonable model to assume is one of computers that can think and learn similarly to a human, just with much higher input bandwidth for simple facts and a bit limited free reasoning ability. That assumption may go a bit to far, but I think it's still a much better model than thinking of it as an improved spreadsheet. Look, for example, at google translate - that is in essence a computer learning the translation between languages from humans, without actually being taught anything explicitly. It's no big magic, and yet the results are quite good overall.
And private companies aren't the only ones playing that game, of course, the NSA has a similar power dynamic, and the borders aren't all that clear anyhow, of course, as any data piles that private companies hold tend to also attract intelligence agencies and the like.
But let me try and show some concrete examples of where personal information is or could be used in order to gain power:
In the political arena, I think that gerrymandering is a good example: Parties use known correlations between personal information they know and voting behaviour in order to increase their chances of winning the election (instead of making the election as representative as possible, which would make a functioning democracy).
Or a company could primarily fire people who have predispositions for certain illnesses that could reduce their efficiency later on or they could right from the beginning only hire those who are not affected. If the pool of workers is large enough, that reduces costs. And as a company needs to be competitive, it actually might not even be able to avoid it once competitors start such a practice.
Similarly for insurance companies: From the perspective of the insurance company, their financial goal in a competitive market is to get rid of any customers that will cost them more money than they pay, so whatever data they are able to get their hands on, they probably will try to use for predictions, and as above they will be forced to do so once some competitor does it. From the perspective of society as a whole, though, insurance is particularly important for those expensive cases, as that stabilizes the social structure, while an insurance industry that only insures people who don't need an insuracnce is essentially worthless for society.
Or suppose a totalitarian leader gets elected. No easier way to make sure that noone challenges your power than to rank the social graph of your country by number of edges and putting in jail anyone who is too well-connected. One important historical case of this type was after Nazi Germany had invaded the Netherlands, where they had all the census information on Hollerith punch cards, including a person's religion. That information was collected without any bad intentions in mind, and yet it ultimately was used for easily finding the jews to kill them.
Or remember the case of Daniel Ellsberg? Nixon's people broke into his psychiatrist's office in order to try and steal his file, so they could use information from it to discredit him.
Also, how about the use of cellphone location data for drone strikes against people who have had no chance to defend themselves in a court, what the US government calls "targeted kilings"?
Well, I guess that's enough for now ... ;-)