Mckinsey ran a comprehensive study of nearly 800 different jobs in the United States, ranging from CEOs to fast food workers. Between these roles, they found 2,000 individual work activities, and assessed them against 18 different capabilities that could potentially be automated. In their analysis, they found that 45% of work activities representing $2 trillion in wages can already by automated based on proven technology that currently exists. A further 13% of work activities in the U.S. economy could be automated if the technologies used to understand and process human language were brought up to the median human level of competence.
WHY IT MATTERS: this report from Gartner surprisingly is very useful as it provides a framework and target reference architecture for machine learning and AI. Very well done and useful. The link is for 2018 update but the original 2017 report is available for free here: https://www.gartner.com/binaries/content/assets/events/keywords/catalyst/catus8/preparing_and_architecting_for_machine_learning.pdf