There's a cost to automating any task: the time spent designing and testing the system. So if it doesn't make economic sense to automate a task, then employment will occur instead.
People are also conservative by nature, so society tends to shun radical new ideas. That conservatism is why it's usually necessary to start your own company in order to put such ideas into practice: almost no one believes it'll work until it's shown to work, so it's rare for big companies to make huge changes to their established business model.
Most new companies aren't huge successes on the scale of Google because success is distributed according to a power law. There are only a few big winners, and big winners are the only companies that can influence a whole society. Therefore it seems mathematically unlikely that most jobs will disappear to automation because most companies won't automate unless they're economically forced to. So even if some work could theoretically be done by a machine, most companies will choose to employ people right up until competitive pressure forces them to embrace automation or go out of business. And that competitive pressure is necessarily infrequent due to the rate of new and hugely-successful companies being infrequent.