The impact of computerisation and automation on future employment

By Hugh Durrant-Whyte, Lachlan McCalman, Simon O’Callaghan, Alistair Reid and Daniel Steinberg. This piece is an extract of the chapter The impact of computerisation and automation on future employment from the CEDA report Australia’s future workforce? published in June 2015. The full report can be downloaded at www.ceda.com.au

Forty per cent of jobs in Australia have a high probability of being susceptible to computerisation and automation in the next 10 to 15 years according to recent National ICT Australia research.

Administration and some services jobs are particularly susceptible, as are regions that have historically associated with the mining industry. Jobs in the professions, in technical and creative industries and in personal services areas (health for example) are least susceptible to automation. We need to look into the future to understand where we now stand.

Providing insight into the qualities of future jobs and what makes these uniquely human and not automatable helps inform us of what we should currently be doing to educate and train the next generation.

Trying to understand and quantify the impact of emerging technology on jobs in areas such as artificial intelligence, robotics and machine learning is key to predicting future employment.

The methodology and initial data used in this study of Australia is based on the much-cited paper by Frey and Osborne, which examined this same problem for the United States (US) and more recently the United Kingdom (UK).

Specifically, the starting point for the analysis is a probabilistic labelling of job categories as 'automatable' based on an expert's subjective view of where technology is leading.

In the next decade or two, the largest impacts will come from the automation of intellectually and physically routine jobs for which machine intelligence and robotics already play a large role. Also important will be the use of automation to substantially increase productivity, and thus reduce employment, in many non-routine occupations.

An important trend is the automation of analysis roles for which machine learning is especially effective and the increasing use of data is key, rather than creative roles for which computer algorithms have yet to make a mark.

There has already been notable work in automation in Australia, especially in the mining industry, replacing human operators on trucks, loads, drills and trains with autonomous vehicles operated by computer from remote location. Increasingly this extends to other occupations in mining including geologists, surveyors and other routine operations.

A second important area will be increasing automation of routine office jobs by computers and machine-learning algorithms. This has already come a long way in the past 30 years: from word processing to automated stock trading. However, current machine-learning algorithms are taking a larger share of what were once perceived skilled jobs or roles around customer engagement. This includes occupations such as legal clerks, market research and sales.

These examples show that the scope of computerisation is growing, especially in analyst-type roles. These probably have the largest impact on data presented in Figures 1 and 2, particularly in a service-centred economy like Australia.

A third important area are jobs which, while not always routine, will see increased productivity (and therefore relatively less employment for the same output) through application of robotics and machine-learning algorithms. Health is an increasingly significant area likely to be impacted – through automation in clinical data and predicted diagnostics (analysis roles), to robotics assisting in areas from surgery to nursing and from hospital logistics to pharmaceutical dispensary. Other examples include banking and legal advice that involve a qualified professional but where data and analysis play a large role, and where most but not all work is routine. Figure 1

Figure 1 shows the estimated probabilities of the susceptibility of jobs to computerisation and automation in Australia, segmented into notionally high, medium and low probability sectors. Notably, 40 per cent of current jobs have a high probability (greater than 0.7) of being computerised or automated in the next 10 to 15 years. This is a lower figure than that for the US (50 per cent) – we believe due to smaller numbers of workers in the service sector – and is comparable to the UK. Figure 2

There are a number of clear messages from the data. First, potential job losses are polarised: Jobs in administration and sales (and many service areas) will disappear, while jobs in the technical professions and personal services will remain. Second, many of those jobs remaining are characterised by non-routine thinking and especially high levels of originality and creativity.

The rate of job impact from automation is substantial and will likely cause significant social adjustment as new jobs and new ways of working emerge. However it is important not to be too pessimistic; this data clearly takes no account of the creation of new jobs and new opportunities in fields that will be opened up by this digital disruption.