Guest author: Flo Falkenhagen (email)
AI-created image of a London taxi driver observing automated vehicles providing transport services
Automation is one of the defining technological developments of modern economies but concerns about machines replacing human labour are not new. From mechanisation during the Industrial Revolution to contemporary robotics, artificial intelligence (AI) and autonomous systems, technological change has repeatedly altered how work is organised and who performs it. This discussion is particularly pertinent as self-driving vehicles become a commercial reality. Automation can substitute for human labour, complement workers’ capabilities, or reorganise jobs so that people perform different tasks. It can raise productivity, reduce costs and create opportunities, but it can also reduce demand for particular occupations, put pressure on wages and make work less secure for those whose tasks are most exposed to change (Autor, 2015).
The key question is therefore not simply whether technology creates or destroys jobs. Automation redistributes work, income and bargaining power between firms, consumers and workers. New employment may eventually emerge, but that does not guarantee that displaced workers can access it on comparable pay or conditions. These issues are especially visible in transport, where AI-enabled systems and automated vehicles may affect drivers directly while also changing logistics, supervision and the organisation of work. This paper considers the history and distributional effects of automation, explores AI’s implications for working practices, and applies these lessons to transport and the UK’s developing regulatory framework for automated vehicles.
Historical development of automation
Automation is a long process of technological change rather than a phenomenon that began with computers or AI. During the Industrial Revolution, machinery in industries such as textiles expanded productive capacity while changing the organisation and pace of work. Workers resisted when mechanisation threatened skilled employment, earnings and control over production. The Luddite protests are an enduring example. Their significance is not that workers were simply opposed to progress, but that productivity gains can create immediate, concentrated costs even when wider economic benefits emerge later. Mokyr, Vickers and Ziebarth (2015) show that anxiety about technological unemployment has accompanied major technological changes for centuries. Innovation can increase output and create new economic activity, yet the absence of permanent economy-wide unemployment does not mean that transitions are painless. Workers may not regain their previous income or status, and communities reliant on a shrinking industry may face prolonged disruption. The time taken for new opportunities to emerge, and workers’ ability to reach them, are therefore central to evaluating technological change.
Mass production in the late nineteenth and early twentieth centuries further demonstrated that automation can transform jobs without immediately eliminating them. Assembly lines standardised tasks and allowed machinery to perform an increasing share of production, while workers remained essential to the system. Computerisation later extended this process into routine cognitive work. Autor, Levy and Murnane (2003) found that computers were particularly effective at substituting for routine manual and cognitive activities that could be expressed through explicit rules, while complementing work involving non-routine problem-solving and complex communication. The content of a job can change substantially even when the occupation itself survives (Autor, 2015).
Acemoglu and Restrepo (2019) describe two forces shaping the result: a displacement effect, when technology takes over tasks previously performed by people, and a reinstatement effect, when new tasks emerge in which human labour retains an advantage. The balance between them matters more than the technology’s capability in isolation. Contemporary AI extends automation into tasks involving prediction, language, pattern recognition and some forms of decision-making. However, technical possibility is not the same as full occupational replacement: adoption costs, reliability, regulation, customer preferences and the design of work all shape what happens in practice.
Who benefits, and who bears the costs?
Businesses can be among the most immediate beneficiaries of automation because it can reduce costs, improve consistency and increase output. Using evidence from 17 countries, Graetz and Michaels (2018) found that industrial robots contributed to productivity growth and lower output prices. They attributed approximately 0.36 percentage points of annual labour-productivity growth to increased robot use. They did not find a statistically significant fall in total employment, although the employment share of low-skilled workers declined. This illustrates why aggregate employment figures can hide substantial changes in the position of particular groups.
Consumers may benefit from lower prices, greater convenience and more available services. Workers’ experiences depend on the tasks they perform and whether their skills complement new technology. People undertaking predictable, codifiable activities may face reduced demand or pressure on wages, while others may gain from new responsibilities and higher productivity. The relevant questions include not only how many jobs remain, but what they pay, where they are located, whether their conditions are acceptable and whether workers can realistically move into them. That transition is rarely frictionless. New jobs may require qualifications, retraining, geographical mobility or working patterns that are difficult for displaced workers to access.
Wang et al. (2023), examining alternative occupations for US heavy-truck drivers, identify possible pathways but find that many offer lower wages than truck driving; some areas exposed to potential losses also lack enough suitable alternative employment. A theoretical job match is not equivalent to a realistic replacement for a worker’s income, location or family circumstances. Regional labour-market conditions, training costs and the quality of available jobs all shape the distributional consequences of automation.
Artificial intelligence and the quality of work
AI makes these questions more urgent because its applications extend beyond physical machinery into administrative, analytical and communication tasks. The UK Government’s 2026 assessment of AI and the labour market notes that around 70 per cent of UK workers are in occupations containing tasks that AI could potentially perform or enhance. This is a measure of exposure, not a prediction that 70 per cent of jobs will disappear; many exposed roles also involve judgement, interpersonal responsibility or physical context that may make AI more complementary than substitutive. The assessment emphasises that the effects will depend on adoption, how organisations redesign work, worker adaptation and policy responses (Department for Science, Innovation and Technology and AI Security Institute, 2026).
Employment is only one part of the issue. AI can alter the pace, autonomy and monitoring of work. Algorithmic systems may allocate tasks, set schedules, recommend actions, record performance and rate workers. Kellogg, Valentine and Christin (2020) show how algorithmic control can give managers new ways to direct and evaluate labour, while also prompting worker resistance. In transport, software that optimises routes, assigns deliveries or measures driving performance may reduce empty mileage and improve safety; it can also intensify workloads, make performance targets less transparent and limit workers’ discretion over how they do their jobs. If decisions affecting pay, shifts or continued employment are based on opaque models, workers may struggle to challenge errors or discriminatory outcomes. AI’s impact should therefore be assessed through job quality as well as job numbers.
Productivity gains can support better pay, shorter hours and safer work, but these outcomes are not automatic. They depend on who controls the technology, whether workers are consulted, what data is collected, how decisions can be appealed and whether savings are shared. Employers and regulators should distinguish tools that assist workers from systems designed principally to intensify performance monitoring or reduce labour costs. Training should also cover the practical use of AI and the limits of automated recommendations, rather than treating digital skills as a one-off qualification.
Automation and the future of transport
Transport is particularly exposed because many activities involve repetitive, rule-based or predictable tasks. Automated railway signalling, aircraft autopilot systems, logistics software and driver-assistance technologies are already embedded in transport operations. Automated and autonomous vehicles extend this trajectory into occupations where human driving remains central. Potential benefits include operating efficiencies, reduced exposure to some forms of human error and improved mobility for people unable to drive themselves. The scale and distribution of those benefits, however, depend on where and how the technology is deployed and how firms, workers and public authorities respond.
Meyer et al. (2017), using the Swiss national transport model, found that automated vehicles could substantially increase accessibility across municipalities. Their modelling also suggested that increased accessibility could encourage urban sprawl and make public transport less attractive outside dense urban areas. These are modelled projections rather than observations of a fully automated transport system, but they underline that the effects extend beyond individual jobs. If automated travel reduces the perceived cost or inconvenience of car journeys, additional traffic and greater dependence on private or shared vehicles could offset some benefits. Accessibility gains must therefore be considered alongside land use, congestion and the viability of public transport.
Freight transport provides a clear example of why forecasts need to consider tasks, timing and business models. Groshen, Helper, MacDuffie and Carson’s influential 2018 report, Preparing U.S. Workers and Employers for an Autonomous Vehicle Future, estimated that autonomous cars and trucks could directly displace between 1.3 million and 2.3 million US workers over 30 years across the scenarios they modelled. The largest effects in their projections occurred decades into the transition, not immediately, and the authors stressed that the costs could be concentrated in particular communities or become more severe during a recession.
These are US scenario estimates, not a forecast for Britain, but the analysis offers a useful framework for assessing the risks and planning ahead (Groshen et al., 2018). The report’s specific analysis of truck drivers is especially relevant. It cautions against assuming that automating the act of driving automatically eliminates every truck-driving job. Drivers also handle freight, interact with customers, complete paperwork, comply with safety requirements and operate non-truck equipment. The report notes that these tasks may be harder or less economical to automate, and that the division of work between driving and non-driving tasks will depend on the business model and regulatory choices.
Some systems may replace a driver on certain routes; others may change the role, with people undertaking remote monitoring, local transfers, loading, delivery or exception handling. The future of truck driving is therefore not predetermined by the technical capability of a vehicle alone. Groshen et al. also argue that the broader gains from automation are not a sufficient response to the losses suffered by particular workers. New jobs may emerge in vehicle technology, maintenance and other sectors but they may be in different places, require different qualifications or offer different wages. The report consequently recommends early engagement between employers, workforce-development institutions and other stakeholders, supported by better labour-market information and policies to mitigate transition costs.
This is a crucial lesson for the UK: workforce planning should begin before displacement accelerates with realistic alternative occupations and training linked to actual vacancies rather than generic promises of reskilling. Research on other countries reinforces this task-based approach. Wang et al. (2023) show that alternative occupations for US truck drivers vary in pay and geographic availability. Bratanova et al. (2026), studying Australia’s road-freight sector, identify 17 occupations with relatively transferable skills but report trade-offs between wage levels and job availability. Bus and coach driving and earthmoving plant operation appear comparatively viable pathways, while some delivery and forklift roles may be easier to access but offer lower median wages.
These findings are not direct evidence of UK outcomes, but they demonstrate why employment-transition plans need to be specific to local labour markets, qualifications and workers’ circumstances. Passenger transport also shows that effects can be felt before a whole occupation disappears. Yu et al. (2026), analysing Baidu’s Apollo Go robotaxi service in Wuhan, estimate that its introduction was associated with a 10.9 per cent short-run reduction in traditional taxi drivers’ average daily income. Their driver survey also reported longer working hours, greater job stress and lower job satisfaction. This is evidence of pressure on a directly competing occupation in one city and over a particular observation period; it should not be treated as a universal estimate for all taxi markets. Nevertheless, it illustrates how competition from automated services may affect earnings and working conditions before full driver replacement occurs.
The UK regulatory framework and workforce implications
The UK has taken an ambitious approach to regulating automated vehicles. The Automated Vehicles Act 2024 establishes the legal basis for authorising self-driving vehicles and creates a framework for safety principles, legal responsibilities and ongoing oversight. The Government has described the framework as one of the most comprehensive of its kind anywhere in the world (Department for Transport, 2025). In May 2026, the Automated Passenger Services permitting scheme opened for applications to support small-scale pilot deployments, while the full regulatory framework remained on track for 2027 (UK Parliament, 2026). The implementation programme is developing the secondary legislation, authorisation processes and in-use regulatory arrangements needed for deployment (Automated Vehicles Act 2024; Centre for Connected and Autonomous Vehicles, 2025). Its focus on safety, accountability and continued monitoring is important because automated vehicles operate in public spaces where failures can have serious consequences. However, a safety and liability framework does not, by itself, guarantee a fair workforce transition.
In the UK, the potential impacts extend beyond heavy-goods-vehicle drivers to taxi and private-hire drivers, bus and coach workers, delivery drivers and people employed in dispatch, fleet operations and vehicle support. The timing and scale of change will vary by use case: motorway freight, fixed-route services, urban robotaxis and last-mile delivery face different technical, commercial and regulatory conditions. Jobs may be reduced, redesigned or created in remote operations, maintenance, cybersecurity, software support, vehicle testing, incident investigation and compliance. These emerging roles will not necessarily be located where driving jobs are lost, nor will they automatically offer equivalent pay or security.
The UK could use its regulatory development to make workforce impacts more visible alongside safety performance. This need not mean treating job preservation as a reason to block safe innovation. It could mean monitoring changes in employment, earnings, hours and job quality; requiring transparent reporting from major commercial operators; consulting unions and workforce representatives during deployment; and publishing regional assessments for sectors likely to be affected. Publicly supported skills programmes should be designed with employers and local training providers and evaluated against job placement, pay and retention. Where disruption is concentrated, targeted career advice, transition income support and place-based investment may be more effective than expecting workers to relocate or retrain at their own expense.
AI also makes continuing oversight necessary after a vehicle has been authorised. Decisions about remote supervision, incident response, performance monitoring and the use of worker data can shape the quality and accountability of new jobs. Clear responsibilities, adequate staffing, paid training and meaningful routes to challenge automated management decisions should form part of good practice. The UK’s regulatory framework provides an opportunity to pair technical assurance with systematic learning about social and labour-market consequences, rather than measuring success solely by the number of vehicles deployed or the safety performance of the technology.
Managing the transition
The history of automation supports neither the claim that technology inevitably produces mass unemployment nor the assumption that its gains will be shared automatically. Outcomes depend on which tasks are automated, which new tasks emerge, how work is reorganised and how institutions distribute the costs and benefits. AI intensifies the debate because it can affect cognitive as well as physical tasks, while algorithmic management can change the experience of work even when a job remains.
In transport, automated vehicles may improve efficiency and accessibility but also put pressure on the earnings, security and autonomy of drivers and other workers. Groshen et al.’s analysis shows why early preparation matters: large-scale effects may take time to arrive, but the costs of transition can be lasting and concentrated. Policymakers and employers should identify credible alternative occupations, link retraining to real vacancies, monitor regional exposure and provide support where displaced workers cannot readily regain comparable employment. They should also ensure that AI-enabled monitoring and work allocation are transparent, proportionate and open to challenge.
The UK’s automated vehicle framework creates a timely opportunity to track these outcomes alongside safety and to consult workers as deployment develops. Automation should therefore be understood not as a binary switch from human-operated to computer-controlled systems, but as a long-term process of occupational and social transformation. Technology shapes what is possible; choices made by governments, businesses and workers shape who benefits and who bears the costs. The policy challenge is to secure the benefits of safer, more efficient transport while ensuring that those affected have realistic and dignified routes into future employment, with decent pay, working conditions and a voice in how new systems are used.
References
Acemoglu, D. and Restrepo, P. (2019) ‘Automation and new tasks: How technology displaces and reinstates labor’, Journal of Economic Perspectives, 33(2), pp. 3–30. Available at: https://doi.org/10.1257/jep.33.2.3.
Automated Vehicles Act 2024, c. 10. Available at: https://www.legislation.gov.uk/ukpga/2024/10/contents.
Autor, D.H. (2015) ‘Why are there still so many jobs? The history and future of workplace automation’, Journal of Economic Perspectives, 29(3), pp. 3–30. Available at: https://doi.org/10.1257/jep.29.3.3.
Autor, D.H., Levy, F. and Murnane, R.J. (2003) ‘The skill content of recent technological change: An empirical exploration’, Quarterly Journal of Economics, 118(4), pp. 1279–1333. Available at: https://doi.org/10.1162/003355303322552801.
Bratanova, A., Mason, C., Evans, D., Schleiger, E., Grimberg, E., Walker, G., Pham, H. and Bulled, K. (2026) ‘Truck drivers and automation: A methodology for identifying and supporting workforce transition in the Australian road freight sector’, Transport Policy, 176, article 103907. Available at: https://doi.org/10.1016/j.tranpol.2025.103907.
Centre for Connected and Autonomous Vehicles (2025) ‘Automated Vehicles Act implementation programme’. GOV.UK, 26 February. Available at: https://www.gov.uk/guidance/automated-vehicles-act-implementation-programme.
Department for Science, Innovation and Technology and AI Security Institute (2026) ‘Assessment of AI capabilities and the impact on the UK labour market’. GOV.UK, 28 January. Available at: https://www.gov.uk/government/publications/assessment-of-ai-capabilities-and-the-impact-on-the-uk-labour-market/assessment-of-ai-capabilities-and-the-impact-on-the-uk-labour-market.
Department for Transport (2025) ‘Automated Vehicles Act 2024 implementation’. Written statement to Parliament, 10 June. Available at: https://www.gov.uk/government/speeches/automated-vehicles-act-2024-implementation.
Graetz, G. and Michaels, G. (2018) ‘Robots at work’, The Review of Economics and Statistics, 100(5), pp. 753–768. Available at: https://doi.org/10.1162/rest_a_00754.
Groshen, E.L., Helper, S., MacDuffie, J.P. and Carson, C. (2018) ‘Preparing U.S. workers and employers for an autonomous vehicle future’. Upjohn Institute Technical Report No. 19-036. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research. Available at: https://doi.org/10.17848/tr19-036.
Kellogg, K.C., Valentine, M.A. and Christin, A. (2020) ‘Algorithms at work: The new contested terrain of control’, Academy of Management Annals, 14(1), pp. 366–410. Available at: https://doi.org/10.5465/annals.2018.0174.
Meyer, J., Becker, H., Bösch, P.M. and Axhausen, K.W. (2017) ‘Autonomous vehicles: The next jump in accessibilities?’, Research in Transportation Economics, 62, pp. 80–91. Available at: https://doi.org/10.1016/j.retrec.2017.03.005.
Mokyr, J., Vickers, C. and Ziebarth, N.L. (2015) ‘The history of technological anxiety and the future of economic growth: Is this time different?’, Journal of Economic Perspectives, 29(3), pp. 31–50. Available at: https://doi.org/10.1257/jep.29.3.31.
UK Parliament (2026) ‘Written question HL796: Steps to support the deployment of self-driving vehicles and autonomous passenger transport services in the UK’. Answered 23 June 2026. Available at: https://questions-statements.parliament.uk/written-questions/detail/2026-06-09/HL796.
Wang, S., Mack, E.A., Van Fossen, J.A., Medwid, L., Cotten, S.R., Chang, C.-H., Mann, J., Miller, S.R., Savolainen, P.T. and Baker, N. (2023) ‘Assessing alternative occupations for truck drivers in an emerging era of autonomous vehicles’, Transportation Research Interdisciplinary Perspectives, 19, article 100793. Available at: https://doi.org/10.1016/j.trip.2023.100793.
Yu, Z., Wang, J., Zuo, T., Alm, J., Li, X. and Li, X. (2026) ‘Robotaxis reduce taxi drivers’ income’, Humanities and Social Sciences Communications, 13, article 629. Available at: https://doi.org/10.1057/s41599-026-07366-x.