Slate.com had an insightful article by Zeynep Tufekci about how software automation (using artificial intelligence via software algorithms) is risky and bad for customer service.
In the article Ms. Tufekci describes the employee-to-customer ratios at the largest web companies:
A similar dynamic dominates policies of social networking platforms—and you only need to look at the employee/user numbers to understand that it could not be any other way. Facebook has about 2,000 employees to 750 million users; Twitter 600 employees to about 100 million users. That’s only three human employees per million users for Facebook and about six for Twitter. Google is larger, with about 30,000 employees, but an enormous portion of the 2 billion netizens use many of its services every day. It probably has about 50,000 users per employee, but for a broader range of services.
With such ratios, the business model becomes to push work onto the user—for example, have the users flag/report what they consider inappropriate content—and then automate the rest.
This explains in part why it is frustrating to use Google’s free products when something breaks down: there is literally no one to help you! You must service yourself by combing through support forum posts and hopefully lucking upon a clear solution (or at least unresolved explanation).