Can a robot do your job?

Find out how much of your job can technically be automated.

Loading interactive...
* * *

Recent advances in artificial intelligence have made it technically possible to automate many tasks that previously could only be done by a human.

But the complex nature of modern work also means that people don’t just do one thing in their jobs. Workers, and especially white-collar workers, complete a wide variety of tasks at work every day.

This interactive calculator, based on data from the McKinsey Global Institute, gives an indication of how the future of work will change: instead of destroying entire jobs and creating completely new occupations, for the most part AI and automation will simply change what activities people focus on in their work.

Nearly every occupation has some tasks that can be automated, but fewer than 5 per cent of occupations can technically be fully automated using technology that is currently available, according to McKinsey data.

Jobs that can be fully automated
  • Meat packers
  • Plasterers and stucco masons
  • Ophthalmic lab technicians
Jobs with no automatable activities
  • Historians
  • Mining roof bolters
  • Clergy

Even for activities that can be technically automated, whether or not it would actually happen is dependent on a number of other factors: the cost of deploying robotic solutions, the economic benefits of doing so, the supply and demand of human labour, and regulatory and social acceptance.

These factors will differ across industries and geographies, and the changes, McKinsey estimates, will take place over decades.

“It takes real effort to integrate technologies and develop solutions that will solve particular business problems,” says Michael Chui, a partner at McKinsey and one of the authors of the report.

“For the many technologies we studied from history, they usually took somewhere between 8 to 28 years between the time of their commercial availability to their eventual plateau in adoption.”

Methodology

This interactive calculator was made using data provided to the Financial Times by the McKinsey Global Institute. It tells you only whether an activity could technically be automated using technology that is currently available. Just because an activity could technically be automated does not mean that someone has made a product or solution to do so.

McKinsey’s team, using US Bureau of Labor Statistics data, deconstructed 820 occupations into their constituent work activities. There are around 2,000 unique activities. McKinsey then assessed, for each activity, what combination of 18 different performance capabilities were required to do that activity.

For example, the job of retail salesperson is made up of activities such as ‘Greet customer’, ‘Demonstrate product features’ etc. Greeting customers requires capabilities such as ‘sensory perception’, ‘social and emotional sensing’, and ‘natural language generation’.

McKinsey then considered what level of performance of that capability is required, based on how humans currently perform those activities. They also assessed whether existing automation technology could achieve the same level of performance. An activity is considered technically automatable only if the answer is yes for all of the capabilities required to do that activity.

For ease of navigation, we have grouped the 820 occupations into 97 job groups, under 23 job categories.

Since some activities require different capabilities depending on the context of the occupation, this means some job groups contain activities that can technically be automated for some occupations but not others in the same group.

For example, for the “Forest, Conservation, and Logging Workers” group, the activity of “Inspect equipment or facilities to determine condition or maintenance needs” can technically be automated for logging equipment operators, but not for forest and conservation loggers.

We have marked these activities as “sometimes”, and have not included them in the overall count of activities that can be technically automated.

Illustrations by Maria Raymondsdotter