How susceptible are jobs for computerisation?

Paul A. Kirschner & Mirjam Neelen

What do we know about the future of our jobs[1]? Do you ever feel uneasy or anxious that yours might disappear before you do? Or do you sleep soundly because you’re convinced things will not turn that quickly?  There is loads of stuff out there on this topic. Many are from think tanks, (inter)national centres, agencies, and professional consultancy firms (for example in accountancy). They’re often times very woolly and contain a high proportion of blah dee blah.  However, one of them is an exception to the rule. Carl Benedikt Frey and Michael A. Osborne’s excellent report from 2013 on “The future of employment: How susceptible are jobs to computerisation?” is a true gem.

The authors examine how susceptible jobs are for computerisation. First, they describe changes in society and technological progress. Then, they estimate for 702 jobs how vulnerable they are for computerisation and how all this came about. Based on these estimates, they research the expected impacts of future computerisation on the (American) labour market. Their estimates take the relationship of the probability of the occupation’s computerisation, wages, and the required educational attainment into account.

Watch out, technology is taking over!

Benedikt Frey and Osborne analyse the history of technological revolutions (this is the terminology they use!) and employment particularly well. They start with the Enlightenment all the way through the 20th century where first mechanisation and next computerisation mainly took place from routine physical tasks (for example, William Lee’s stocking frame knitting machine in 1589 and machines in cotton manufacture around 1800) and later on routine cognitive tasks, such as assembly lines at Ford Motor Company that take over continuous-flow processes in 1913. The authors go on to distinguish technological revolutions of the 21st century. They discuss computerisation of (non-) routine physical task and (non-) routine cognitive tasks. When doing so, one can distinguish the following four quadrants:


They observe a change from computerisation of jobs that were mostly carried out based on routine tasks (both physically / mechanically and cognitively), to jobs that also require non-routine physical / mechanical and cognitive skills. This transition has been caused by the growth and maturation of artificial intelligence, data analytics, and machine learning.

They state for example that:

[T]he rapid pace at which tasks that were defined as non-routine only a decade ago have now become computerisable is illustrated by Autor, et al. (2003), asserting that: “Navigating a car through city traffic or deciphering the scrawled handwriting on a personal check – minor undertakings for most adults – are not routine tasks by our definition.” Today, the problems of navigating a car and deciphering handwriting are sufficiently well understood that many related tasks can be specified in computer code and automated (p 15).

The authors explain this by the availability of ‘big data’ in combination with the progress that machine learning has made and is still making. This combination enables us to design algorithms that keep improving themselves compared to human achievements. Indeed, they argue that the:

[C]omputerisation of cognitive tasks is also aided by another core comparative advantage of algorithms: their absence of some human biases. An algoritm can be designed to ruthlessly satisfy the small range of tasks it is given. Humans, in contrast, must fulfill a range of tasks unrelated to their occupation, such as sleeping, necessitating occasional sacrifices in their occupational performance (Kahneman, et al., 1982). The additional constraints under which humans must operate manifest themselves as biases. Consider an example of human bias: Danziger, et al. (2011) demonstrate that experienced Israeli judges are substantially more generous in their rulings following a lunch break. It can thus be argued that many roles involving decision-making will benefit from impartial algorithmic solutions (p 16)

One example that illustrates this process is the computerisation of diagnostic tasks in healthcare. For example oncologists at Memorial Sloan-Kettering Cancer Centre use IBM’s Watson computer to provide chronic care and cancer treatment diagnostics. Watson holds knowledge from 600,000 medical evidence reports, 1.5 million patient records and clinical trials, and 2 million pages of text from medical journals. These are all used for benchmarking and pattern recognition purposes. Watson is used in order to diagnose and develop a treatment plan with the highest probability of success because the computer is able to compare each patient’s individual symptoms, genetics, family and medication history, and so forth.

Another example shows how computerisation is entering the domains of legal and financial services. Law firms now rely on computers that can scan thousands of legal briefs and precedents to assist in pre-trial research. Symantec’s Clearwell system is a frequently cited use case; it uses language analysis to identify general concepts in documents, it can present these results graphically, and it has proved to be capable of analysing and sorting more than 570,000 documents in two days.

We’re better than they are

What are the current limitations of computerisation of jobs? Benedikt Frey and Osborne’s response to this can be summarised in 3 concepts; that is social intelligence  (for example, being able to, in real time recognise human emotions and respond adequately to them), creativity (coming up with ideas and artefacts that are unusual), and perception and manipulation tasks (for example, complex perception tasks in an unstructured work environment and then being able to respond to that environment). The figure below shows the probability of computerisation of jobs based on these three “bottleneck variables”.


Top & Flop 5

Of the 702 jobs that the authors have analysed, the top five that will remain accessible for employees on the job market are:

1    Recreational therapist

2    First-line supervisors of mechanics, installers, and repairers

3    Emergency management directors

4    Mental health and substance abuse social workers

5    Audiologists

The flop five; that is the most likely jobs to be computerised and hence disappear from the labour market are:

698    Insurance underwriters

699    Mathematical technicians

700    Sewers (hand)

701    Title examiners, abstractors, and searchers

702    Telemarketers


Autor, D., Levy, F., & Murnane, R. J. (2003). The skill content of recent technological change: An empirical exploration. The Quarterly Journal of Economics, 118, 1279–1333.

Frey, C. B., & Osborne, M. A. (2013). The Future of Employment: How Susceptible are Jobs to Computerisation? OMS Working Papers, September 18. doi:

[1] Paul A. Kirschner is currently doing a fellowship with NIAS (Netherlands Institute for Advanced Study in the Humanities and Social Sciences), which is financed by NSvP (Dutch Foundation for Psycho Technique). The fellowship focuses on (mainly vocational) education and studies two main questions. First, how can youth and workers be optimally prepared for the unknown and unpredictable labour market of the future and second, what is the role of information skills (that is, primarily information literacy and information management) in that education (