Robots and AI eating into job market for predictable and routine work

29 August 2016

Robots and early AI are already proficient enough to displace vast tracts of predictable physical work and data process tasks. The effect of automation may see more and more tasks taken away from humans, across a wide range of professions, either creating more time for different tasks – or seeing a reduction in FTEs. Management of people remains an area with a low automation potential – given current technologies.

Automation has the potential to free humanity from the fetters of repetitive, physically demanding and, often, unpleasant work. The loss of such work, particularly where few skills are required, may have undesired effects – such as mass unemployment, and misery of another kind. The first industrial revolution, while in the long term leading to improve human prosperity – particularly in the west – came with mid-term costs, as generations of low skilled and artisan workers were side-lined.

As a new form of the industrial revolution, focused primarily on AI and robotics, appears to be rapidly approaching – questions are being asked about the social and economic consequences. The end of drudgery, may also mean increased unemployment, income inequality and suffering. While ways forward are being considered, including models like basic income, the extent to which AI and robotics will eventually affect workers is being explored by a range of institutions and organisations. McKinsey & Company's latest thought piece on the topic looks specifically at the technical capability of the technology to displace traditionally human tasks. 

The analysis highlights that the rise of robotics and machine learning is more nuanced than a clear cut – robot replaces human – narrative. It is likely that robotics and AI will take specific tasks away from human operators. This may mean that humans are open to doing more, or, if there is no more that they can or need to do, that one person takes on more of a different task, replacing another. The article is based on an analysis of more than 2,000 work related actives in the US across 800+ occupations, derived from the US Bureau of Labor Statistics and O*Net.

Feasibility of replacing human tasks with new automation technology

As it stands, the current line of robotic and AI technologies, will begin to heavily affect around 51% of all work time in the US. Much of this time is spent on data collection, data processing and predictable physical work.

The analysis suggests that predictable physical work, performed in an environment that is relatively stable, such as in food service and accommodations, manufacturing and retailing, is highly susceptible to automation. Particularly manufacturing is open to transformation, with around a third of all work in the environment considered predictable. Given the relative ease of automation, at current technological advancement, up to 78% of tasks could be replaced.

It is not merely low-level production work that is open to this form of automation, however, as data processing and data collection tasks too are open to becoming robotised. The firm calculates that 47% of retail salesperson’s activities have the technical potential to be automated – although this is far below the 86% of mid-level professional activities such as bookkeepers, accountants, and auditing clerks that can be replaced by an automated system.

Sector specific automation of tasks

The research also sought to identify which positions have the most potential for automation. Educational services and management are the areas least likely to be heavily affected by the current best technologies. Educational services, in particular, has low levels of the three most easily automated tasks, data collection, data processing and predictable physical work. Managers can expect some of this work to be automated, particularly those pertaining to data collection and processing. Construction and agriculture, which involve large amounts of unpredictable physical work, will see some automation – although the nature of the tasks currently protects workers in the segment from replacement.

The firm notes that the analysis is focused primarily on the ability for current technologies to displace tasks – not that such tasks will necessarily be replaced. The cost-benefit analysis is not itself part of the analysis, and, in certain situations, workers remain cheaper and more plentiful than the capex for a robotic system. Even here, however, changes in the cost of technologies are rapidly advancing, with in some domains, particularly RPA, the costs of an automated system may far outweigh that of employment. New technologies, particularly those that allow for improved natural language interactions – in which robots can understand and respond – would, according to the firm, open vast new tracts for possible automation.