As far as I understood from previous coverage and reading the privacy policy, the point is that e-mails are mined even if ads are turned of for the domain. The resulting profiles are then used for showing contextual advertisements in services that are not in Google Apps (e.g. Google search and Google+). Google's lawyers have also admitted that this is true. IANAL, but reading the privacy policy and the ToS for Business accounts, it seems to be the same there.
Of course, you can completely opt out of interest-based ads, both on Google services as on Google ads across the web. But I assume that profiles are still built, if not used.
A related problem is that persons sending e-mail to a GMail address (which could be hidden behind a non-gmail domain) never consented to the ToS and their e-mails are profiled. To which Google's reaction was: "all users of email must necessarily expect that their emails will be subject to automated processing." [1] IMO there is a difference between scanning e-mail for spam and viruses, and using the content to build a profile of the sender or receiver.
[1] http://www.theguardian.com/technology/2013/aug/14/google-gma...
Also the cusomised ad advertising is an option you can opt out of, so once again I'm not understanding the issue here beyond grabbing some headlines for mistakes that were avoidable on many levels.