I have been doing applied statistics in combination with qualitative research and industry analysis; all pretty traditional project based kind of work since the mid 90s. In recent years as I added in NLP, SQL and web dev style programming skills to the mix I find myself building apps that in many ways automate analytical tasks.
With one product I developed for professional and trade associations I estimated that the average association would require as many as 10 full-time entry level analysts working with the typical Excel, Access and Powerpoint tools that so many of them use in order to even come close to replicating the personalized, graphical reporting that my platform could do for an association in real-time.
After feeling a bit guilty here is what I came to realize. The automation that I am offering lets the association offer a set of services to their members that previously would have never been considered because it was financially unthinkable. So if an association decides to use my platform they won't typically eliminate jobs. Instead they will add really substantial capabilities with an external platform supported service and no additional headcount.
This is just my example, but I think that might be true in many cases that automation of analytics (data science) will mostly add capabilities to organizations and not take away many jobs.