Abstract
This article investigates the job satisfaction and working conditions responses of welders after the implementation of collaborative welding robots in their workplaces. Through semi-structured interviews and ethnographic observations at three sites, it explores themes such as pay, prestige, professional identity, social interaction at work and career prospects. Our research suggests that working experiences and responses to the robot are highly varied among welders, and many individuals were ambivalent about the new technology. We found evidence of improved work experiences, such as improvement in occupation safety, offloading repetitive tasks and the social prestige related to working with a robot, but also drawbacks and anxiety surrounding the adoption, change in professional identity and the possibility of being replaced by robots in the future. Building on previous literature on automation and job displacement, especially on the impact of new technologies on blue-collar workers, we broaden academic debates from questions surrounding labour replacement to delimiting contexts where new technologies enable the retention of workers and the creation of more meaningful work.
Introduction
According to the UK government (Gov.UK, 2022), about one-third (34%) of the workforce in the UK is employed in what are traditionally considered ‘blue-collar’ occupations. These jobs are commonly understood to require significant manual labour and often mechanical skills to operate specialised tools and machinery. Throughout the long history of technology and employment debates, blue-collar workers have been widely predicted to be among the biggest victims of technological change (Snell and Gekara, 2023). The literature on the future of work, particularly economic calculations and statistics, generally highlights the risk of massive job displacement sparked by automation technology. Indeed, these arguments and predictions of the imminent technological overtaking of blue-collar jobs can be traced back many decades. For example, Casey argues in Work, Self and Society after Industrialism (1995: 37) that ‘by the 1980s, a number of occupations and former job categories such as welder, riveter, switchboard operator, salaries clerk, bookkeeper, typesetter (among many others) had disappeared’. In a similar vein, in The End of Work (1995), Rifkin expected that ‘by the mid-21st Century, the blue-collar worker will have passed from history, a casualty of the Third Industrial Revolution’ (p. 140). Such views are still common in more recent studies of Industry 4.0 (Ford, 2015; Frey and Osborne, 2017), and worries about job displacement have broadened to include white-collar jobs as artificial intelligence (AI) is increasingly capable of performing tasks that were once considered too complex for machines (Forbes, 2025; Susskind and Susskind, 2015).
There are good reasons to argue that blue-collar workers are most vulnerable to technological advancement. First, amid the continuous increase in labour costs in advanced economies such as the UK (BCIS, 2023; Trading Economics, 2025), capital cannot generate profit when the wages of workers are too high; therefore, the owners of capital are keen to reduce their costs by introducing robotics and automation technology into their workplaces (Fleetwood, 2012). Second, despite a plethora of examples of technology failure, the assumption of technology’s superiority over labour in performance efficiency and reliability, especially in non-cognitive and routinised tasks, is still widely held in society; one need not be a Marxist scholar to accept that capital, by nature, constantly seeks productivity advances (Marx, 1973). Third, new technology in the workplace can often fragment the labour process and, consequently, reinforce managerial control and authority over workers, which is sometimes used by management as a countermeasure to the (organic or political) solidarity developed in the labour process. These characteristics are clearly manifested in labour-intensive industries such as steelwork or manufacturing, in which this study is contextualised. The latest statistics (Office for National Statistics, 2025) showed that in 2023, 14% of the firms in the UK’s manufacturing sector adopted robotics, with 64% of the firms adopting other ‘specialised equipment’ for automation. The adoption rate is expected to be higher today, as the UK government has been promoting these ‘Smart Machines’ to drive economic growth (Smart Machines Strategy 2035) (see also Department for Science, Innovation and Technology, 2025).
Despite these motivations, we should not uncritically accept that the replacement of blue-collar work by new technology is inevitable. In fact, research predicting the demise of blue-collar work has been criticised for its overly technological-determinist theoretical orientation. In work and employment studies, the dominant view understands the effects of technologies to be socially shaped (Edwards and Ramirez, 2016). Researchers such as Barley (2020) have long advocated for more social and political examination and theorisation to understand the institutional and organisational influences on the implications of technology on work, the dynamics of power and the forces of resistance. This call has yielded a series of research studies that reveal the augmentation properties and constraints of artificial intelligence and advanced robotics in today’s economy (e.g. Liu and Hayton, 2024, 2025). Despite the literature pointing out the theoretical and empirical weaknesses of the reductionist thinking of job displacement, ‘technological determinism is still here and unlikely to disappear’ (Wyatt, 2008: 175).
Against this background, we place the experience of blue-collar workers at the centre of our analysis. Blue-collar work can be further divided: jobs that require extensive training and qualifications, and jobs that require only basic skills. Labour studies has a rich tradition of research on the impact of technology on blue-collar work. Under the heavy influence of Marxist scholarship, Braverman (1998 [1974]) and his intellectual followers (and, of course, critics), paid most of their attention to assembly-line factory workers. There is a somewhat surprising lack of qualitative studies of the job satisfaction and working conditions of high-skilled blue-collar professionals, such as welders and fabricators. The lack of discussion in academic literature gives rise to the misconception that welding is a dying occupation in today’s economy, which further influences people’s view of the profession.
This article aims to fill this knowledge gap by examining welders’ job satisfaction and working conditions after the implementation of collaborative welding robots, specifically the one we call RobotW. Drawing on in-depth interviews with 20 stakeholders in the welding industry who have first-hand experience working with RobotW, complemented by ethnographic observations at three sites, we argue that such impact is highly ambivalent, and there are both positive and negative emotions affecting workers’ job satisfaction. In this article, we focus on welders’ subjective experience and their narratives over generalisability. We try to understand their world of work from the richness of lived experience and their narratives, which capture temporality, human emotion, meaning and plot. 1 We contribute to the theoretical debate on automation and blue-collar work through these narratives and experiences, which is a complement to the existing (statistical) studies in the literature (Filippi et al., 2023; Klenert et al., 2022). Moreover, some studies treat blue-collar work as one category of jobs in the labour market; yet researchers (Autor et al., 2003; Frey and Osborne, 2017) argue that the impact of technological transformation on workers varies by the type and level of skills required in their jobs (that is, less skilled workers handling routinised tasks are, arguably, at higher risk of technological substitution); therefore the investigation of welders in this study, as an example of high-skilled blue-collar workers, can provide insights that are different from studying less skilled blue-collar occupations.
After this introduction section, we set up the theoretical background on the technological impact on blue-collar work, particularly welders, by reviewing the literature. We then introduce our research methods, present and analyse our research findings, and finally conclude with a discussion, signposting a direction for future research.
Revisiting the job displacement thesis: Robotic automation, new technology and blue-collar work
The fear of technological disruption has long existed in society, often fuelled by economic predictions such as John Maynard Keynes’s ‘technological unemployment’ thesis (Keynes, 1931), suggesting that automation technologies such as industrial robots will inevitably lead to massive job losses. Indeed, the impact of robotic technology and information technologies (such as computers) on the job market has been much examined in labour economics research over the decades. Researchers are interested in understanding how these new technologies affect employment numbers, as well as productivity, the polarisation of skills and wages (e.g. Acemoglu and Restrepo, 2020; Acemoglu et al., 2023; Fierro et al., 2022; Frey and Osborne, 2017). However, their findings regarding robots and employment are contradictory.
At first glance, robotic automation appears to be destructive toward blue-collar jobs. Brynjolfsson and McAfee (2014) show that automation caused an overall displacement of blue-collar jobs in automotive manufacturing in the US. Similarly, Dauth et al. (2017) conclude that ‘every robot destroys two manufacturing jobs’ (p. 4) in the German manufacturing sector, accounting for almost 23% of the overall decline in manufacturing employment. Studies using aggregate International Federation of Robotics (IFR) data tend to show negative or null correlations between robotisation and employment (e.g. Borjas and Freeman, 2019; Graetz and Michaels, 2018; IFR, 2019). In particular, using data from 27 European countries and the US, Happanala et al. (2023) show that industrial robots have caused a deeper decline in industry-sector employment (for young and less educated workers), if a country has higher union density. Robotisation also appears to have an adverse effect on workers’ wages in the US manufacturing context (Guarascio et al., 2025), although the overall relationship between robotisation and wages is statistically insignificant (Guarascio et al., 2025; Jurkat et al., 2025).
However, other researchers challenge these claims. By studying the steel industry in five European countries, Antonazzo et al. (2024) show that automation and digitalisation of steelmaking do not necessarily substitute workers in the process, as their experience, contextual knowledge and judgement are required to complement the technology. Moreover, Klenert et al. (2022) show that from 1995 to 2017, robotic automation created more manufacturing jobs than it replaced, and countries with high levels of (robotic) automation have been more resilient in the face of ongoing manufacturing decline. In other words, industrial robots are positively associated with total employment in Europe’s manufacturing sector. While these studies are a powerful corrective to the popular narrative of technological job displacement, some blind spots in the debate still require further investigation. For instance, economic studies focus primarily on the number of jobs or skills; there is a lack of holistic, conclusive understanding of how robotic automation affects blue-collar workers’ job satisfaction and working conditions.
Using data from the World Robotics Survey and the European Working Conditions Survey, Antón et al. (2023) argue that robotisation has a negative impact on the quality of work as it increases work intensity, but there is no relevant impact on physical environment or skills and discretion; on the contrary, Gihleb et al. (2022) argue that robotisation leads to a decrease in (physical) job intensity in the German context, but no significant effects on workers’ mental health and work-life satisfaction. Job satisfaction and working conditions are important because they directly contribute to workers’ decisions to accept or resist the technology and workers’ well-being. Past studies (e.g. Snell and Gekara, 2023) demonstrate that technology implementation at work is a collective process, and workers can exercise their agency to either collaborate or resist its adoption. Technology implementation in the workplace can be beneficial to workers’ job quality (Peeters et al., 2025); and workers perceiving themselves more susceptible to automation also show more willingness to keep up with technological development and engage in learning activities to upskill and/or reskill themselves (Jansen et al., 2025), making them more resilient to job lost. Statistics (ten Berge and Dekker, 2025) suggest that labour organisation decreases workers’ perceived level of job insecurity related to technological change, as workers’ organisation enables them to give voice during the implementation process or even to push back against the technology adoption. (Cieslak and Valor, 2025). Also, the process by which (skilled) blue-collar workers experience changes in job satisfaction and working conditions due to technology adoption is often neglected. We do not intend to reintroduce the entire intellectual debate of technology and blue-collar work here (Gekara and Nguyan, 2018 provides an excellent summary); instead, this article aims to contribute to filling this research gap by presenting the impacts, warts and all, of collaborative welding robots on welders in the UK.
Moreover, the studies mentioned above mainly consider industrial robots, which are centrally controlled machines with which workers have to interact – evidence shows they can intensify work rhythms and labour effort making monitoring workers’ performance easier, and making workers become more dependent on the pace of work of machines (Weil, 2014). However, the latest version of manufacturing robots, such as the RobotW discussed in this article, are designed to be more flexible and collaborative with humans to handle multiple tasks. 2 These robots are also enhanced by information technology and can be trained by AI (e.g. Huang et al., 2025), which has greater potential to be disruptive. While these studies refer to recent and ongoing trends in the labour market at the time they were published, academic research must keep up with technological development.
Indeed, academic studies on the impact of welding robots on welders over the last decades primarily focus on their skills (Erden and Tomiyama, 2009; van Essen et al., 2008), performance (Erden and Marić, 2011) and productivity (Ferraguti et al., 2023); there is little discussion on how robotic technologies impact their work experiences. Therefore, we anchor our work in an earlier study: Mutch’s (1998) seminal contribution on the impact of technology on welders in the UK.
In order to examine Catherine Casey’s (1995: 186) claim that the welding profession, among others, will become obsolete after automation and information technologies are introduced, Mutch (1998) conducted fieldwork in four factories in the East Midlands, England, interviewing welders to understand the value of human labour in the production process. He discovered that the introduction of welding machines on the shop floor did not lead to layoffs and the substitution of welders. Instead, the machine required more highly skilled workers to set up, monitor and maintain it. Further, he found that human welders have higher flexibility than machines to adapt both to the technical needs of the process (e.g. working with different materials) and to the vagaries of market demands. Therefore, automation neither reduced the need for skills nor replaced workers – much less occupations.
Building on Mutch, this article’s contribution to the literature is twofold. First, we explore the impact of collaborative welding robots on English welders with themes related to work conditions and working experience. Rather than seeing technological impact as a one-way, cause-and-effect association, we highlight the complexity of the relationship between a new technology and the people who operate it, which often is a mixture of both positive and negative experiences. Recognising the subjective experience of work is valuable as it directly contributes to the meaningfulness of work (Bailey et al., 2019). The article’s empirical focus addresses the important question posed by Edwards and Ramirez (2016): ‘when should workers embrace or resist new technology?’ By closely examining workers’ subjective work experience and working conditions, we can understand the ‘direct and indirect . . . intended and unintended’ effects of technological impacts on work (Edwards and Ramirez, 2016: 101–102).
Second, a key question derived from studies comparing new and old robotic technology is: ‘is it really different this time?’ (Wajcman, 2017). Ford (2015) and others firmly believe a fundamental shift in the relationship between workers and machines is rendering workers more vulnerable to replacement. By examining welders and their interaction with new technologies in a similar context, the unique insights generated by this study enrich this debate. Our empirical findings can give a voice to workers and enable unions, management and policy workers to make more informed choices.
Research context and methods
This article draws on an ongoing qualitative study of the developments and changes in blue-collar work and how new technology impacts working conditions and work experience as part of the Pissarides Review into the Future of Work and Wellbeing, the welding robot is a technology strategically chosen to shed light on understanding changes in UK’s manufacturing and construction industries.
Social context
Despite their centrality to the functioning of society, the UK has been experiencing difficulty in attracting and retaining these workers. This labour shortage crisis is highlighted in various policy documents, including the annual report of the Migration Advisory Committee (2024), which urges the government to further investigate the relationship between labour supply and post-Brexit migration policy. The shortage of welders in the UK is among the government’s most critical labour policy challenges. One estimate (Axiom Personnel, 2023) suggests that by 2027, half the UK’s 61,000 welders (Statista, 2024) will have retired, creating over 35,000 new vacancies. Labour shortages can result in project delays and increased costs, and the potential for improved efficiency has incentivised more employers to introduce new robotic technologies into their workplaces. In response, among other things, in 2024 UK initiated a £50 million apprenticeship pilot project to support the growth of 13 selected sectors, including pipe and plate welders programmes (ECITB, 2024). While numerous new apprentices enter the market every year and new welders are college graduates, many of them are equipped with only a basic skillset that is not up to industry requirements. Some businesses have turned to technology to fill the gap. The lack of skilled welders and challenges in recruitment have fostered the development of automation in the welding industry, which can be traced back to the 1990s (Mutch, 1998).
Research methods
Ethnographic approaches and interviews have proven to be effective ways to study the implications of technological change for work and collaboration in the factory environment, especially to capture the subjective experience of work (Brannan et al., 2007). The methodological package deployed in this study strictly complied with the University of Warwick’s guidelines regarding data collection and storage and had been approved by the ethical review committee of the University before data collection began.
The technology we examined in this article, RobotW, is the latest iteration of a collaborative welding robot launched in September 2023. Unlike traditional industrial robots with safety barriers that separate human workers from the machine, RobotW is designed to work alongside human workers on the shop floor. It is controlled via a tablet: a worker sets up the coordinates of the robotic arm movement, either using a simple programming language or manually ‘hand-holding’ the robot’s welding gun. Once the robot is programmed and calibrated, it operates the welding gun repeatedly. The operator simply loads the material to be welded and unloads the finished product. The machine does the welding.
Few machine shops in the UK had installed RobotW when our study began, so recruiting research participants proved challenging. Through the Warwick Business School network, we approached the UK RobotW dealership, who provided a list of buyers. Using a purposive sample method, we approached some buyers of this technology and invited them to participate in this non-commercial research project.
While we made it clear that our research project was funded by the Nuffield Foundation and it is not for profit-making purposes, some participants who initially showed interest in being interviewed stepped back, because they were not authorised to speak for the company and worried about inadvertently revealing privileged information about their business (i.e. national defence, R&D in vehicle manufacturing, structural steel construction). We also rejected potential participants who had purchased RobotW but had not yet put it in operation.
In the end, we interviewed 19 practitioners and one regional RobotW sales representative (see Table 1), all with first-hand experience operating and working with the robot, and many with accredited welding qualifications (in industry terms, they are ‘coded’ welders). During our interviews, they sometimes referred to themselves and their colleagues as fabricators or platers (instead of welders) because of the different roles and tasks involved in their respective workplaces. However, in reality, welding is frequently part of fabrication work; it is less so for platers, but they are often expected to be able to do some basic welding. To facilitate our discussion here, we refer to all of them as welders.
Informant data.
Note: All interviewees are men.
A set of interview questions was developed to capture the everyday work content of these welders: how does the RobotW impact their work, and how do they feel about these changes? The first author was responsible for all interviews, fieldwork, data collection and coding, while the second author provided essential research support and supervision. These interviews lasted 30–60 minutes and were audio-recorded with the interviewees’ consent. All names are pseudonyms. The interview data was then transcribed using the university-licensed MS Word, and codified according to different aspects of job qualities such as pay, working hours, work intensity, social relations at work; these codes then became the themes that guide the discussion in the next section.
While on-site during October and November 2023 to conduct semi-structured interviews at three machine shops, in Sussex, Sheffield and Durham, ethnographic observations were made by the first author. We do not provide a regional comparison of technology’s impact (our colleagues at Institute for the Future of Work have done this, brilliantly), but gathering data in three regions ensures that our data are not corrupted by a biased local contextualisation. Between interviews, the first author spent time on the shop floor, in offices and in break rooms, and had lunch and socialised with workers. With the participants’ consent, photos were taken by the first author to visually document the welders and their labour processes on the shop floor. We believe the photos presented in this article will help reader understand the workplaces that our interviewees and the robots operate in and their working conditions.
Research findings and analysis
Common experiences: Pay and occupational health and safety
We begin our analysis with what is widely considered to be the most important aspect of a job: pay. We are interested in understanding how the introduction of the collaborative welding robot affects welders’ salaries, and also how they do their work. In the UK, according to government data (National Careers Service, n.d.), the average salary for an entry-level welder is £25,000 a year, which can progress to £45,000. The average salary of a welder/fabricator in 2024 is £33,500, which is 9.5% higher than in 2023, making welders the occupational group with the highest salary rise across all professions in the UK’s labour market (Office for National Statistics, 2024). This massive pay rise was primarily driven by the shortage of skilled welders (The Engineer, 2025).
We asked our interviewees whether or not the introduction of a collaborative welding robot in their workplaces has had any direct effect on their salaries. All interviewees responded that there had been no direct impact, except at one machine shop. As Robert explained, in his workplace, Twenty per cent of the company’s profits go into a pot. And at the end of the year, we share it out. So the robot[’s work] will be put into the pot, but it won’t want anything back from it, so it’s gonna actually earn us more bonuses at the end of the year.
All of our interviewees acknowledged that RobotW has improved their occupational health and safety. Minor occupational injuries, such as thumb burns or skin burns, are common in welding operations. However, during robotic welding, because the metal is fixed on a welding table and the welding gun is programmed from the control tablet, the welder can control the operation from the tablet (Figure 1) and be completely hands-off during the welding process. This prevents accidents that cause minor burns (Figure 2).
A welder setting up a RobotW using a tablet. © Hong Yu Liu, 2024.
A welder showing the burn marks on his hands (thumb and little finger). © Hong Yu Liu, 2024.
New ways of work and the changing work experience
Apart from improved occupational safety, welder experiences of the changes brought by the robot vary to a large degree. Most interviewees believed that working with a welding robot increases their work efficiency. Once the robot is correctly set up, it can weld two to three products in a single run. Then the operator gathers more materials and the process is repeated, with the robot completing the task dozens or even hundreds of times. Manual welding is a one-at-a-time operation. RobotW’s advantage is seen to be especially valuable during high-volume, repetitive processes. Some interviewees believed that welders could focus on more ‘creative’, non-routinised welding tasks by offloading these repetitive tasks to the robots. As production manager Chris told us, Sometimes we need 50 of those [products] really quickly, . . . which would be very mundane for manual welding . . . the welders are just going to be working like a robot. It’s just not good for your mind; it’s just not good for your job satisfaction. . . . That’s what your granddad did, and you don’t want to be doing that.
In addition, three welders explicitly stated that they enjoy working RobotW. When we spoke with Charles, a fabricator foreman with nearly 10 years’ experience, about how he feels working with a robot, he said
In a way, it [the robot] takes that job off me, doesn’t it? Well, I just put it in there and then just clean it up afterwards and then put the next one in. So it is a bit easier for me.
So it is a bit monotonous. Does it bore you?
No, not if you are the one who is setting it up. I had the robot set it up yesterday; it was my first call, which I quite enjoyed. Actually, it wasn’t too bad, but I know it [job satisfaction] comes from [comparing with] my last job . . . the company I was at, we did mobile welding on ships. [wry laugh] . . . and for site work, we cannot bring a robot in.
Conversely, Derreck, who is a welding expert with six welding qualifications, holds a completely different point of view. When the interviewer asked if Derrick thinks the introduction of the collaborative welding robot benefits him at work, he replied:
I do not think that anyone is benefited . . . I don’t know. . . . I think that it can quite slow the process down, because, yeah it is a robot. You can press one button and you can weld it, but also, in every job, you still need someone to load the robot up, clip it down into position, press the button, weld it and take it back out.
But if you are welding 2000 pieces of the same part day in, day out, are you not going to get bored?
No – it just gets easier and easier. . . . It doesn’t bother me whatsoever.
Derreck’s viewpoint is shared by Andrew and Colin, who also preferred working on routinised over non-routinised, ‘creative’ tasks. While some people find them mundane, Andrew thinks routinised tasks make his work predictable and manageable: I know I can do 25 pieces daily, and I have one week to finish all these jobs. Say, I could leave at 3 o’clock today, because I know I still have more time than I need to get it done tomorrow.
Not everyone enjoys working with the welding robot. A couple of interviewees told us that the robot has made their work more difficult, especially those involved in preparing material for the welding process. This is because the robot lacks a human welder’s flexibility and, therefore, has far less engineering tolerance. Jason illustrates this point by describing his current project, working on 1000 steel stumps: Because steel doesn’t come perfectly sized. . . . This is fine from a fabrication point of view, but when I give it to the robot, it has a pinpointed starting point and a finishing point. The robot wants everything to be exactly the same and everything perfect. Whereas with things like steel and stuff, there are engineering tolerances. . . . Welders can easily adjust accordingly but the robot isn’t designed to do that.
For this reason, Louis, an engineer and commercial director, described the welding robot as a ‘checker’ to the welder. Welders are under pressure to maintain a pace that does not delay shop floor operations. Suppose a welder gives the robot imperfectly sized materials as they feed it; they will notice immediately that the robot’s weld is inadequate. Manual welders possess engineering tolerance and can make welds that enable a greater range of materials to meet customer expectations. However, RobotW’s shortcomings leads to wastage of both materials and time. This shortcoming also causes social tension, because managers blame the welders who prepared the materials, even though RobotW’s engineering intolerance is the problem.
Prestige, social interaction and professional identity
During our field visits, many welders were clearly excited to see a new welding robot installed in their workplace. The new robot seemed to bring prestige to these welders; on many occasions, we saw welders photographing or video-recording themselves working with their robots and then sharing these photos and videos with their friends and families.
Having the latest equipment provided some welders with a sense of satisfaction. Even so, other interviewees told us that the real pride of being a welder comes from ‘welding it manually’ (Figure 3). As John explained,
When you are doing welding manually, and then you’ve done it, and you think, ‘Oh wow’.
How does that differ if it is done on a robot?
It is a completely different experience. When welding something with my hands, we had to do it aesthetically, so it looked nice and was pleasing. When I see a nice weld done by a robot, it’s like, ‘yeah, it is a nice weld, but I haven’t done it’. It’s not the same feeling . . . you don’t get the same emotional attachment.
A welder doing manual welding next to a RobotW. © Hong Yu Liu, 2024.
Indeed, being able to build things with their hands is a primary source of job satisfaction for many of the welders we spoke to. For example, when we asked what he likes most about his job, Colin, a 73-year-old welder with over 50 years’ experience, said, ‘the only thing I don’t like is clocking in and out’, suggesting that despite all the hardship inherent in welding (such as the flame, the noise and the high temperature), he found his work enjoyable. Andrew’s answer is even more straightforward: ‘The satisfaction comes from putting something together, and it actually looks like something it is meant to be.’
Another source of job satisfaction mentioned by some welders is the social aspect of their work on the shop floor. The introduction of a robot inevitably has affected social relationships at work.
Working with human welders comes with its own benefits that cannot be replaced by the welding robot:
Would you prefer to work with a robot or with human colleagues?
I’ve worked with a lot of colleagues and stuff like that, and it’s been a good time, we all enjoy doing the welding. . . . I have met quite a few people . . . [names of his colleagues] . . . so yeah, it is more fun being with someone and having a little bit of connections. Work can take a bit longer to do sometimes, but I think it’s still a healthier environment and [especially] when you are getting on with everyone and everyone likes the same thing. Yeah, everyone just gets one better. . . . And I think working on your own with just a robot can kind of play on your mind a little bit.
To enhance social engagement, one machine shop had a full-fledged entertainment room, with table football, video game consoles, drink machines and even drones for employees to play with during breaks. Another had a coffee room and some entertainment opportunities. The third had a coffee machine.
These days however, the shrinking workforce makes hiring experienced welders a major challenge for many machine shops. Commercial manager William told us he would much prefer his workplace to invest in a welding robot over hiring a welder, because there is a high risk that a new hire will not be good enough for the job as a good welder needs years to develop his craftmanship. A welding robot can provide the consistency needed for a large-volume production, and it outperforms inexperienced human welders. But it does not provide the sense of community among welders.
Apart from the lack of emotional attachment to the machine and the missing human interaction, welding with a robot has led some welders to reflect on and question their professional identity. When the interviewer asked about drawbacks associated with using RobotW, Brian replied: I guess the negative drawback would be that my job description is a fabricator/welder and, I mean, at that point, I would sort of [re]classify myself as a machine operator because it [the robot] is still a machine. I sort of see myself like that . . . but I still want to be a fabricator. . . . [Programming the robot] is just putting some numbers in and moving it from one point to another; it is not exactly hard to do per se.
Skills, training and career prospects
The conversation about how easy/difficult a welding task is and its impact on a welder’s identity led to a question about what makes a welder professional, that is, their qualifications and skills, both of which concern the importance of training.
In the structural steel industry, which six of our interviewees are working in (Figure 4), every single welding operation has to be done according to the Standard Welding Procedure set by a qualified construction engineer for quality assurance, and a qualified welder must perform this welding work. Recently, this constraint expanded to allow a welding robot approved by the governing body – British Constructional Steelwork Association – to do this work.
A fabricator preparing structural steel. © Hong Yu Liu, 2024.
Whether or not a qualified welder is needed to operate a robot is an important question related to the career prospects of these welders and the future of the industry. Take our interviewee Justin as an example. He is hired as a machine operator at a workshop to operate RobotW, and describes his everyday work as ‘setting speed and feed’.
As mentioned, the UK government has established an apprenticeship project. However, as collaborative welding robots are a relatively new technology, we found that none of our interviewees who recently completed their apprenticeship phase had experience working with welding robots prior to RobotW during their current job. While welders can learn how to operate a welding robot on the job, this training led some to reflect on the training programmes and qualifications they received from their colleges:
If I knew I was going to be doing robotic welding and it would have only taken six months to [learn], then I wouldn’t have done a fabricating and welding course for two years.
But do you think the two-year course makes you a better operator when you work with the machine? Can you do a better job with the machine?
Yes. I think the manual side of learning . . . when I am welding, I can control my own welding pool, I could do it upside down, I could do it inside a closed area. There are a lot of perks to doing it manually, but it’s just the speed of doing it is different compared to a welding robot. . . . So, for a robot to get into a tight gap or anything like that, you might mess five up before you get it right. Whereas if you added it manually, I could do it right the first time. So that’s my skill on that part.
We glean two insights from James’s response. First, the government’s apprenticeship initiative seems not to be perfectly tailored to contemporary industrial needs. Second, for James and others like him, RobotW is likely to hinder the development of welding skills on the job.
Meanwhile, not every welder we interviewed found on-the-job training easy. When asked if he has received enough training to operate the robot, Jason replied, Now I have to keep my head into it and do it every day. Chances are, if you told me go down there and programme it now, I’ve forgotten half of it. I jot down notes, but if you are not doing it all the time, I don’t think I know what to do. . . . but I have had the training and if you gave me some time to play around it, I could deal with it.
Using a tablet to control the welding robot is particularly difficult for Colin (Figure 5), a welder with more than 50 years’ experience:
I’ve only had one go on it. I wasn’t sure what to do. They wanted me to have a go to see if I like it.
Would you like to receive further training?
No, . . . I will leave it for somebody else. Computers are not my bit. So I would rather let somebody else come in and do it.
So you are not computer literate?
No.
Colin, a welder with more than 50 years’ experience, was introduced to the first author during his fieldwork. © Hong Yu Liu, 2024.
Colin’s experience is just one example of how difficult it can be for an ageing workforce to keep up with robotic technology. According to the Welding Institute, the average age of qualified welders in the UK is 55+ (EngEPA, 2024); new technologies impose a barrier that directly impacts how welders can contribute to the operation. However, the relationship between age and the value of the manual welding skillset is not straightforward. While Colin understandably believes manual welding is still valuable and sees himself working for this company until he turns 80 in a few years, Jack, an industry veteran who did his apprenticeship in the 1970s, believes that the value of manual welding is depreciating and welders ‘must understand the working of the machine’ to stay relevant.
Despite their differences, all interviewees agree that while a qualified welder is not necessary to operate the robot (Figure 6), knowledge and experience in manual welding helps them operate the robot better and realise its full potential. Commenting on this topic, Charles said:
He [Justin, a plater without a welding qualification] has been asked to weld, and I’ve been told to help him out sometimes and show him how to go about it. . . . Look, if he said it had holes in it or had a bit of impurities in it, Justin wouldn’t understand what they are, whereas I would. So, then I would be able to detect that and say ‘that’s wrong and needs sorting out’.
Would you say working with the robot helps develop your skills?
It has a little bit, yeah. Because I’ve been at a few companies and . . . I guess if I were to go somewhere else and say that I am trained in robotic welding, then that’s a benefit.
A RobotW in operation without human supervision. © Hong Yu Liu, 2024.
Lastly, what does the introduction of the welding robot mean to the future of welding?
Industrial relations literature suggests that unions play a crucial role in protecting workers’ interests when technological changes are introduced to a workplace. However, none of our interviewees is a union member, and none works at a unionised workplace. Louis, who is a construction engineer and owner of a construction steel machine shop, said that I am just glad that our guys are not involved in the unions. I think if they were, you know, we would have had difficulties buying that [robot]. . . . I think unions are one of the biggest challenges for companies to implement [change]. They want to stifle innovation and say it is all about preserving jobs, but it doesn’t preserve jobs at all. It just puts companies out of business because they can’t remain competitive.
We reached out to The Workers Union and to Unite Construction, Allied Trades and Technicians to enquire about their position on collaborative welding robots, but got no reply. While Louis believed that some (unionised) welders oppose the introduction of the welding robot and see it as a threat to their jobs, our interview findings show that (non-unionised) welders have a variety of perceptions regarding a displacement effect. We acknowledge that our data cannot verify whether or not union membership has any effect on welders’ attitudes toward the robot. Nevertheless, our findings signal a complicated relationship between technological advancement and career prospects. Most welders we interviewed believe, for various reasons, their jobs will not be totally displaced in the future. One has to deal with the flexibility of manual welding, as Chris explained: With a few customised brackets and that sort of thing, it is just not worth using it [RobotW]. It takes a little bit longer to programme. So, in that case, a welder would look after those and also do short bracket repairing something like that, which is a lot quicker and easier.
Petros pointed to the robot’s technical limitations: There will always be a need for a welder, and the robot cannot completely replace a welder, in my opinion. . . . A robot cannot do pipe [welding]. They cannot do much for oil rig positions. There is always going to be a need for manual welders, especially site welders.
However, five of our interviewees believe that in the long term, their jobs will be totally wiped out by robots because, at some point, ‘everything is going to be automated’ and welders will become someone who ‘loads and unloads’ (Jack). Welders might become machine operators, someone to ‘look after the robots, to maintain and keep them running’ (Jason). Another reason to be pessimistic is that despite all the recruitment efforts, there are not enough newcomers to sustain the industry and, therefore, machine shops must increase their reliance on robots. When Colin was asked what he thinks about the future of the workforce, he replied, ‘I see the future of the welding fabrication industry shorting of good skilled labour. It’s a shame, really. I should have somebody under me to teach him.’
Discussion and conclusion
Regarding the question, is it really different this time? That is, does RobotW impact worker experiences differently than earlier automation technologies? At first glance, our answer seems to be ‘no’. Some of our findings mirror those presented nearly three decades ago (Mutch, 1998), such as the value of human flexibility and the importance of having a skilled welder to improve the machine’s performance. This is also in line with more recent studies (e.g. Antonazzo et al., 2024) that highlight the importance of experience, contextual knowledge and judgement required to complement the technology. However, our interviewees showed higher degrees of uncertainty when talking about their career prospects, the possibility of being replaced by a robot, and the future of the welding industry. This could be due to the fact that none of our interviewees works at a unionised workplace, as, in principle, unions can moderate the impact of robotisation and the risk of technological displacement (Dupuis and Massicotte, 2025; ten Berge and Dekker, 2025).
Our main findings enrich the dominant labour economics literature on automation and blue-collar work by paying more sustained attention to job satisfaction and working conditions. Each of our interviewees expressed a striking degree of ambivalence regarding RobotW. They recognised the technology’s potential, but did not hesitate to underscore its shortcomings. Past studies (e.g. Antón et al., 2023) investigating the relations between robotisation and job quality highlights how robotisation can lead to increasing work intensity and consequentially more health problems at work, but this is not observed in our study. One possible explanation is that RobotW is a collaborative robot that enable welders to have more control on the pace of work. This helps welders to retain their autonomy and should be considered a positive trait for their well-being at work. The adoption of RobotW in the three worksites has changed the way welders work and consequentially their work experience, but there is no conclusive impact observed in relation to other aspects of job quality, such as wages and skills. This echoes meta-analyses in the economic literature (Guarascio et al., 2025; Jurkat et al., 2025).
Moreover, we found no evidence of variation by worker category: welders of all ages, regions and skill levels expressed similar views on the changes brought about by the welding robot, both positive and negative. Notably, all emphasised that their salaries and working hours have not been affected (except for one worksite, where the robot contributes to the bonus), and occupational health and safety have been improved. Because of the ambivalence, our position is that neither the power of capital nor the technology itself can determine whether certain technology choices should be embraced or resisted (Edwards and Ramirez, 2016).
Although our sample size is small, the range of experiences documented here supports a more optimistic account, compared to doom-filled estimates that, inevitably, many blue-collar workers will be displaced as technology advances (Ford, 2015; Frey and Osborne, 2017). This suggests that technology, by itself, is not a determining factor affecting how workers experience technological change. Therefore, we need to widen our academic attention from the technological unemployment debate to incorporate how, when and why technology use might retain and create more ‘meaningful work’. We must confront broader theoretical and empirical questions of how technology can contribute to better work conditions and working experiences in the labour market. Our findings suggest that welding robots have a huge potential to improve worker satisfaction and self-fulfilment. However, individual-level ambivalence reminds us that the relationship between technology and meaningfulness at work is nuanced. Much existing research suggests that skill level is another determining factor in the extent to which workers benefit from the implementation of a new technology (e.g. Hötte et al., 2023).
As Petersen et al. (2023) argue, the importance of human capital (i.e. education, experience, training) can significantly affect how workers experience automation at work. In a similar vein, our study shows that workers in the same occupational group can experience different benefits (and drawbacks) from technologies, as they are differently affected by factors such as previous work history, changing work content, social relations, pride and prestige, etc.
Our findings remind us not to accept simple and reductionist narratives of new technologies and work experiences, as the evidence points to a rather complicated relationship: workers can be both champions and victims of technological change.
Lastly, we must acknowledge some limitations to this study. First, our data were collected during the early adoption phrase in the three worksites, and it is possible that some of the positive and negative experiences will be neutralised when workers become more familiar with the technology. The long-term effect of robotisation on workers’ job satisfaction could be insignificant, as shown in Gihleb et al.’s (2022) longitudinal analysis. Second, the three worksites in this study are small and medium-sized enterprises (SMEs), the impact of robotisation could be different in larger manufacturing sites with a different organisational structure and work culture. We are also aware of that they are (early) adopters of the technology: Acemoglu et al.’s (2023) study shows how robot adoption could have a different effect on workers in the same industry for adopters and their competitors. How welders working at non-robotised worksites are affected by the robotisation of the industry is a topic deserving to be explored. Third, our study is technology specific, but we restrained our findings from arriving at overly general conclusions about the welding industry, as other welding robots could have completely different characteristics and impacts on workers’ working conditions and work experience. These are possible directions for future research.
Footnotes
Funding and acknowledgements
This research is part of a larger study, The Pissarides Review of the Future of Work and Wellbeing, which was funded by the Nuffield Foundation Strategic Fund (SFS /FR-000022922). The authors acknowledge the significant support from the Institute for the Future of Work, the project Principal Investigator Professor Sir Chris Pissarides (LSE), and our Co-Investigators, Anna Thomas, MBE (IFOW), Dr Abby Gilbert (IFOW), Professor Jolene Skordis (UCL), Dr Jonathan Clark (Imperial), Professor Maurizio Barahona and Dr Bertha Rohenkohl (Our World in Data).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.