The human impact of automation: Lessons from Amazon

Amidst a wealth of evidence of poor working conditions inside the warehouses Amazon calls ‘fulfilment centres’, questions have been raised about whether some of these ‘bad jobs’ might soon be replaced by machines. At the very least, it might be expected that the increasing use of robots would reduce the physical strain on human workers, some of whom have reported walking over 10 miles per shift.

However, members of the Work Futures Research Group at Nottingham Trent University, who reviewed over 500 academic and non-academic sources of information about Amazon’s warehouses, found that so far, automation of the fufilment process is associated with not only a growing workforce, but also worsening conditions for workers – a cause of grave concern for the future.

Amazon is important, as an example of current trends in automation, because its international reach and growing market share place huge pressure on its competitors to imitate its methods. In addition, the number of workers employed by Amazon is growing: 500,000 employees were added to a worldwide total of 1.2 million in 2020 and further expansion has been announced since then.

Following its acquisition of Kiva Systems in 2012, Amazon has deployed over 200,000 Kiva robots in locations across the world. Because the use of these robots, in processes like stowing and picking, has eliminated the need for transportation of goods from one area of the fulfilment centre to another, whether by labour power or by mechanical conveyor, their introduction into the process of fulfilment is associated with a new relationship between workers and their means of production, a new system of work, that comprises, in addition to the worker, five essential pieces of equipment:

  • mobile shelving units called ‘pods’;
  • a grid of QR codes, known as ‘fiducials’, on the floor of the robotic field;
  • a workstation fitted with scales, lasers, sensors, cameras, and touch screens;
  • robotic drive units equipped with movement sensors that can lift the pods and carry them to the workstation;
  • a software programme that directs and coordinates the movement of the robots so that hundreds can be in operation simultaneously and without collision.

Similarly, the introduction of the Pegasus robotic drive unit in 2018, which is intended to increase the speed and accuracy of sorting packages by post code, is associated with a new system of work, but one that involves a new work role, that of flow control specialist, who is responsible for monitoring the operation of the robots, including their traffic patterns, identifying major congestion spots, and adjusting the inbound and outbound package volume, in line with fluctuations in customer demand.

Other automatic machine tools that Amazon has introduced include the robotic pallestiser, which carries loads of up to 165 kilograms by means of mechanical grippers; the Robo-Stow, which also lifts heavy pallets; the SLAM, which is used for weighing and addressing packages and flagging up potential errors in labelling for a human operator; and the CartonWrap, which builds a box around an item, seals it with glue and labels it, at a rate of between 600 and 700 boxes per hour – four to five times faster than a worker can.

Although the introduction of the robotic drive units has eliminated the need for workers to walk long distances when stowing and picking, robotic stowing and picking remain demanding processes because they require movements of the human body which put it under stress – for example, bending downwards to reach bins on lower levels of the pod or stretching upwards to reach bins on higher levels. When journalist Alan Selby worked as a picker alongside Kiva robots in the Tilbury fulfilment centre, in 2017, his fitness tracker showed that, for one shift, his level of exertion was equivalent to walking at least 10 miles.

Moreover, the introduction of robotic stowing and picking is associated with increases to the productivity targets for workers and thus an increase in the speed of the work process. For example, the reported targets for manual stowing and picking at Amazon’s Swansea fulfilment centre, which opened in April 2008, were 90 and 110 units per hour, respectively. By contrast, the reported target for robotic stowing at Amazon’s Eastvale fulfilment centre, which opened in March 2018, was 300 units per hour, while the reported target for robotic picking at Amazon’s Staten Island fulfilment centre, which opened in September 2018, was 400 units per hour. It is unsurprising, therefore, given the increasing pace of work and the risk of developing musculoskeletal disorders and stress that this creates, that automation within Amazon’s warehouses has been associated with increased rates of injury. For example, a US-based investigative journalist working for the non-profit organisation Reveal found that the rate of serious injury in newer warehouses operating with robotics was more than 50% higher than in older warehouses operating without robotics.

Regulating human labour to fit in with fast-paced automated elements of the fulfilment process also places a mental strain on workers. In a survey of UK-based fulfilment centre workers by the Organise workers network, 57% of respondents said that they had become ‘a lot more anxious’ since they started working at Amazon. The main cause of such anxiety appears to be the threat that workers will be the subject of disciplinary action, if they underperform in their role. For example,

  • a former Amazon worker at the Etna warehouse in Ohio told The Daily Beast that he ‘was constantly fearful that he would receive citations’ for not being able to ‘stay on pace’;
  • Vincent Tortora told The Street that, when he worked as a stower at Amazon’s Robbinsville fulfilment centre in New Jersey in 2017, he was afraid of being punished (‘written up’), if he did not scan merchandise quickly enough;
  • in an article on working conditions in Amazon’s fulfilment centres in Logic, a stower at Amazon describes, in a post on Reddit, feeling anxious about not being able to meet performance expectations, saying ‘I can’t get better no matter how hard I try’;
  • John Burgett writes that the three categories of performance evaluation introduced in 2016 were ‘more threatening’ than managers made them out to be, and that the coaching and write-ups that managers would use, if workers were not able to meet performance expectations, were ‘tactics to apply daily pressure, always leaving the employee on unsure footing.’

Managers’ threats of disciplinary action toward workers who underperform rest on the knowledge that workers fear losing their jobs. For example,

  • Business Insider reported Amazon fulfilment centre workers being in a constant state of anxiety about losing their jobs, if they did not meet their performance targets;
  • the US-based fulfilment centre workers who spoke to The Atlantic said that they were fearful of losing their jobs, if they did not meet Amazon’s performance expectations;
  • a worker at Tilbury who responded to the second wave of the Organise survey said that workers were ‘fear[ful] for their jobs’ and that an ‘Amazon fulfilment centre is the most terrify[ing] place to work’;
  • according to International Business Times, ‘Amazon warehouse employees … contend they’re told by Amazon and outsourced managers to meet productivity goals designed to be unattainable for most in an effort to keep them in a perpetual state of insecurity about their continued employment’.

But what also makes production workers feel anxious is the level of surveillance to which they are subjected. For example,

Such surveillance is integral to the automation of the fulfilment process: it is the result of managers’ attempts to keep workers moving at the pace of the machines they are working with and to prevent them from trying to resist the demands of their employer or assert their own needs.

Assessment of the prospects for further automation power must also consider the relative cost of labour power compared to that of machines. Given the rapid turnover of its workforce, Amazon is currently facing a recruitment crisis in some countries and is responding by offering seasonal workers sign-up bonuses.

Yet the state of crisis in the international capitalist economy and the decline in social welfare contribute to a ready supply of disposable labour that can be obtained for low wages, mitigating against full automation of the process of fulfilment. It is therefore clear that automation in itself offers no easy removal of the bad jobs created by Amazon.

The full report on which this article is based has been published here.

Tom Vickers and Dominic Holland are members of the Work Futures Research Group at Nottingham Trent University.

Image credit: Frederic Legrand – COMEO