Will robots take manufacturing jobs? What the evidence on automation, productivity and human capital shows, 2015–2026

Abstract

1. Introduction

Every wave of factory technology has produced the same headline, and the current wave has produced it twice: for industrial robots in the 2010s and for AI and humanoid machines since 2023. The concern is not frivolous. Manufacturing still directly supports about 760,000 jobs in the UK's advanced-manufacturing subsectors alone and contributes more than £82 billion of gross value added a year (Department for Business and Trade, 2025), the sector as a whole ranked eleventh in the world by output in 2025 (Make UK, 2025b), and a wide body of forecasting work has put anything from a quarter to a half of jobs "at risk" (Frey and Osborne, 2013; IMF, 2024; ILO, 2025).

Yet forecasts of exposure are not measurements of outcome. Since 2015 there has been enough data to ask a narrower question: in the countries that actually installed robots at scale, what happened to manufacturing employment, wages and productivity, and what does that tell a UK business owner deciding whether to invest, retrain or wait?

This paper offers a UK-focused reading of that record. Section 2 reviews the theory and evidence, Section 3 the data and calculations, Section 4 the findings, Section 5 their meaning, what is new about 2025–2026 and the implications for UK SMEs; Sections 6 and 7 give limitations and conclusions.

2. Literature review

2.1 Theory: tasks, not jobs

The modern framework treats production as a bundle of tasks, some done by people and some by machines. Autor (2015) argued that automation "substitutes for labor" but also complements it, raises output, and increases demand for the problem-solving and adaptive tasks that machines cannot do; commentators, he suggested, systematically overstate substitution and ignore complementarity. Acemoglu and Restrepo (2019) sharpened this into three named channels. A displacement effect occurs when capital takes over tasks that labour used to do; it always lowers labour's share of value added and may reduce labour demand even while productivity rises. A productivity effect works the other way, because cheaper output raises demand for everything else. A reinstatement effect occurs when new tasks are created in which people have the comparative advantage; it always raises the labour share. The net effect on jobs is not a matter of theory but of which channel is larger in a given place and period.

Autor et al. (2022) supplied the long view on reinstatement: most current employment is in job titles that did not exist in 1940. Their warning is that since 1980 "the demand-eroding effects of automation innovations have intensified" while the augmenting effects have not, so the reassurance that new work always arrives is weaker than it looks.

2.2 Robots and employment: the country studies

Graetz and Michaels (2018) used IFR data for seventeen countries from 1993 to 2007 and found that robots added roughly 0.36 percentage points a year to labour productivity growth, raised total factor productivity, lowered output prices, and "did not significantly reduce total employment", although they reduced the employment share of low-skilled workers. That is a reallocation result.

Acemoglu and Restrepo (2020) reached a darker conclusion for the United States. Comparing commuting zones by their exposure to robots, they estimated that one additional robot per thousand workers lowered the local employment-to-population ratio by about 0.2 percentage points and wages by about 0.42 per cent, with no offsetting gains elsewhere in the same local economy. The US, in their reading, is a displacement story at the regional level.

Dauth et al. (2021) ran the closest comparison for Germany, a country with far higher robot density than the US. Using administrative records for 1994–2014, they found that each robot displaced roughly 1.7 manufacturing jobs, but that these losses were "fully offset by new jobs in services", so total employment was unaffected. Strikingly, incumbent workers at automating plants were more likely to keep their jobs, because they moved into new tasks within the same firm; the displacement fell on people never hired rather than on those in post.

2.3 Robots and firms

Firm-level studies show the mechanism. Koch, Manuylov and Smolka (2021) tracked Spanish manufacturers over 1990–2016: adopters raised output by 20–25 per cent within four years, cut the labour cost share by 5–7 percentage points and grew net employment by about 10 per cent. Acemoglu, Lelarge and Restrepo (2020) found the same in France: the 598 adopters among 55,390 firms held a fifth of manufacturing employment and expanded their headcount, but "at the expense of competitors", so the industry-level association was negative. Robots do not so much destroy jobs as move them from non-adopters to adopters.

2.4 Exposure forecasts and the AI turn

A separate literature estimates how much work could be automated. Frey and Osborne (2013) put 47 per cent of US employment at risk from computerisation. The OECD (2023) noted that AI outputs had become "almost indistinguishable from that of humans" in many areas. The IMF (2024) estimated that almost 40 per cent of global employment, and about 60 per cent in advanced economies, is exposed to AI, of which about half might be harmed and half helped. The ILO (2025), using a finer task index, found 25 per cent of global employment and 34 per cent in high-income countries exposed, while stressing that full automation of whole jobs "remains limited". The World Economic Forum (2025), surveying more than 1,000 employers across 55 economies, expected 170 million jobs created and 92 million displaced by 2030, with nearly 40 per cent of required skills changing and 77 per cent of employers planning to reskill.

McKinsey Global Institute (2024) put the share of European work hours that could be automated by 2030 at about 27 per cent, requiring up to 12 million occupational transitions. Its 2026 follow-up (McKinsey Global Institute, 2026) estimates that 58 per cent of current European work hours could in principle be automated with existing technology, 44 per cent by software "agents" and 14 per cent by robots, and that even in manufacturing 71 per cent of the projected value to 2030 comes from agents. Brynjolfsson, Li and Raymond (2023) supplied the first rigorous workplace evidence on generative AI: 14 per cent more issues resolved per hour among 5,179 support agents, and 34 per cent for novices. Both results matter for factories, whose office and planning functions are where the software lands first.

2.5 The UK policy literature

The Made Smarter Review (Maier, 2017) argued that faster adoption of industrial digital technology could be worth £455 billion to UK manufacturing over a decade, raise productivity by more than 25 per cent by 2025 and add a net 175,000 jobs. Make UK (2025a) placed the UK 24th in the world for robot density. The Advanced Manufacturing Sector Plan (Department for Business and Trade, 2025) responded with up to £99 million for Made Smarter Adoption, £40 million for Robotics Adoption Hubs and £29 million a year to 2030 for innovation.

3. Method

This is desk research using secondary data only.

Robot statistics. IFR press releases for World Robotics 2024 and 2025 give installations, operational stock and density (robots per 10,000 manufacturing employees) for 2022, 2023 and 2024 (IFR, 2024a; 2024b; 2024c; 2025a; 2026). The UK density figure for 2023 is taken from the IFR World Robotics 2024 UK chapter as cited by the Tony Blair Institute (2024), because the IFR's public release for the UK gives installations and stock but not density. Density depends on the employment denominator: the IFR cut China's figure sharply for 2024 after China's National Bureau of Statistics updated its labour data (IFR, 2026).

UK labour market statistics. Four ONS series were used: workforce jobs in manufacturing, seasonally adjusted, June of each year (ONS, 2026a, series JWR7); the Index of Production for manufacturing, chained volume, 2023 = 100 (ONS, 2026b, series K22A); productivity hours in manufacturing, 2023 = 100 (ONS, 2026c, series DK3V); and average weekly regular pay in manufacturing and in the whole economy (ONS, 2026d, series K5DU; ONS, 2026e, series KAI7). The August 2026 labour-market bulletin supplied the latest annual change (ONS, 2026f).

Interpretive framework. The findings are read through the task model of Acemoglu and Restrepo (2019). In its simplest form the change in labour demand can be written as:

Here W is labour demand measured as the wage bill, Y is value added, S is the share of value added paid to labour, R is the reinstatement effect from new tasks, D is the displacement effect from automated tasks, ε collects other influences such as changes in the relative price of capital, and Δ ln denotes a proportional change. The first term is the productivity effect: more output pulls labour demand up. The second term is where robots bite: displacement pushes S down and reinstatement pushes it back up. It is an accounting identity, not a prediction; its use here is to separate the "more output" channel from the "smaller share" channel.

Derived productivity. The ONS publishes manufacturing output per hour as a growth rate; to obtain a level index on the same base as the other series, output per hour was calculated as the output index divided by the hours index:

Here P, Q and H are, for year t, derived output per hour, the manufacturing Index of Production (2023 = 100) and the manufacturing productivity-hours index (2023 = 100), and g is the proportional change between 2015 and 2025, about 16.8 per cent. The 2020 and 2021 values are distorted by furlough, when hours fell far more than output, and are not used for any conclusion.

Annual workforce-jobs values are June observations; percentage changes are the author's arithmetic; no causal estimate for the UK is attempted, because the robot stock is too small and the regional data too coarse for the shift-share designs used in the literature.

4. Findings

4.1 Robot density: the UK is not where the robots are

Figure 1 shows the IFR's 2024 density ranking with the UK's 2023 figure added for scale. The Republic of Korea leads at 1,220 robots per 10,000 manufacturing employees, followed by Singapore (818), Germany (449) and Japan (446); the United States is eighth at 307 and the EU-27 average is 231 (IFR, 2026). The world average was 132, below the 162 published for 2023 and the 151 for 2022 only because of the China revision; before it, the IFR described the global average as having more than doubled in six years (IFR, 2024a; 2024c; 2026). The UK, at 119 in 2023, sits below the world average, below China's revised 166, and at roughly one-quarter of the German level (Tony Blair Institute, 2024; IFR, 2026).

The flow tells the same story. Global installations were about 542,000 in 2024, the second-highest year on record, with an operational stock of 4.66 million; China installed 295,000, 54 per cent of the total, and the IFR expects more than 700,000 a year by 2028 (IFR, 2025a). The UK installed a record 3,830 robots in 2023, up 51 per cent and half of them in automotive, which the IFR linked partly to the expiry of the "super deduction" capital allowance (IFR, 2024b). In 2024 UK installations fell to about 2,500 (IFR, 2025a), roughly 35 per cent lower, against 26,982 in Germany and 8,783 in Italy. The UK's operational stock of 28,831 robots in 2023 was about a ninth of Germany's 269,427 (IFR, 2024b).

4.2 UK manufacturing: fewer jobs, more output, without many robots

Figure 2 indexes the ONS series to 2015. Manufacturing workforce jobs rose from 2.64 million in June 2015 to a peak of 2.72 million in June 2018, then fell to 2.52 million in June 2025 and 2.48 million in March 2026, down 4.4 per cent over the decade and 8.6 per cent from the peak (ONS, 2026a). The latest bulletin records a drop of 81,000 manufacturing jobs, 3.2 per cent, in the year to March 2026 (ONS, 2026f). Over the same decade the manufacturing Index of Production rose from 90.3 to 99.4, about 10 per cent (ONS, 2026b), and hours worked fell from 104.6 to 98.6, about 5.7 per cent (ONS, 2026c). Derived output per hour therefore rose about 16.8 per cent, only 5.5 points of it by 2019; the rest came after the pandemic, as output recovered on fewer hours.

Read through Equation 1, the UK pattern is a productivity effect with a modest displacement effect: output up, hours and jobs down. But the productivity gain is not obviously robot-driven. The UK's robot stock was under 29,000 units in 2023 (IFR, 2024b); the German estimate of 1.7 factory jobs per robot, applied crudely to that whole stock, would explain only a fraction of the decline since 2018. The likelier drivers are ones the ONS series cannot separate: energy costs, trade friction, sectoral mix and the loss of hours during and after the pandemic. Figure 3 summarises the channels through which Table 1 should be read.

4.3 Pay: manufacturing has kept a premium, but a narrower one

Average weekly regular pay in manufacturing rose from £540 in 2015 to £759 in 2025, about 41 per cent, while whole-economy regular pay rose from £453 to £679, about 50 per cent (ONS, 2026d; 2026e). The manufacturing premium therefore narrowed from about 19 per cent to about 12 per cent over the decade, and stood at about 11 per cent in June 2026 (£783 against £703). This is neither the "hollowing out" in which the remaining workers are the higher-skilled and better paid, nor the wage fall Acemoglu and Restrepo (2020) found in robot-exposed US regions. It looks like a sector shedding hours without upgrading its capital stock fast enough to lift pay relative to services.

4.4 Skills and adoption as the binding constraint

The UK evidence on why robots are not installed points at organisational capacity rather than technology. Make UK (2025a) identifies fragmented support, complex funding and a lack of accessible digital-skills training as the main barriers for smaller manufacturers. The sector plan records about 49,000 manufacturing vacancies, and that only 7 per cent of manufacturers were "well-versed" in AI and 8 per cent had implemented it (Department for Business and Trade, 2025). Where support reached SMEs, the evaluated employment effect was positive: 243 beneficiaries of the Made Smarter Innovation programme showed 14–15 per cent higher employment than unsuccessful applicants (SQW, 2025). Whatever its causal weight, that is the opposite sign from the displacement story.

5. Discussion

5.1 Decline of human capital, or reallocation?

The "decline of human capital" hypothesis predicts that as robots spread, manufacturing employment falls, the remaining workers are deskilled or underpaid, and the sector's know-how erodes. The evidence from 2015 to 2026 does not support that reading, for three reasons.

First, the countries with the most robots have not seen the largest job losses. Germany, at 449 robots per 10,000 employees, absorbed its automation through service-sector growth and longer tenure for incumbents (Dauth et al., 2021; IFR, 2026). The UK, with a quarter of that density, has lost manufacturing jobs anyway.

Second, at firm level the effect is reallocation towards adopters. Spanish and French adopters grew output and headcount while competitors shrank (Koch, Manuylov and Smolka, 2021; Acemoglu, Lelarge and Restrepo, 2020). Human capital does not vanish; it moves towards firms that can integrate machines into a production system. That is a reskilling story, and it is why 77 per cent of employers surveyed plan to upskill (World Economic Forum, 2025).

Third, the one clear casualty in every study is the labour share, not employment. Graetz and Michaels (2018), Koch, Manuylov and Smolka (2021) and Acemoglu and Restrepo (2019) all find that automation shifts value added from wages to capital even when jobs are stable. The distributional problem is real; it is not the same problem as mass unemployment.

5.2 The strongest counter-argument

The best case against this conclusion is Acemoglu and Restrepo (2020) combined with Autor et al. (2022). The US commuting-zone results show that displacement can dominate locally for two decades, with real wage losses, and no service-sector rescue arrived in those places. Autor et al. show that reinstatement has weakened since 1980. If the AI wave now automates the very service and office tasks that absorbed Germany's displaced factory workers, the offset that made the German result benign may not be available next time. McKinsey Global Institute (2026) quantifies this risk: three times as much European work is technically automatable by software agents as by robots. A reader who weighs the US evidence above the European, and expects reinstatement to keep weakening, could reasonably conclude that 2025–2026 is the start of something worse than 2015–2024. This paper does not dismiss that view; it notes only that it is a forecast, while the reallocation reading is the record.

5.3 What is different about 2025–2026

Two things. The first is humanoid and general-purpose robots. The IFR's position is measured: humanoids "will complement and expand upon existing technology" rather than replace conventional robots, and mass adoption as household helpers "may not happen in the near or medium term" (IFR, 2025b). The relevant change for a factory is not the human shape but a machine that can be re-tasked by demonstration rather than by an integrator writing code, which would lower the fixed cost that keeps robots out of small-batch production. No dataset yet measures that effect; 2024 installations were still conventional arms (IFR, 2025a).

The second is that automation has moved from the shop floor to the office. The exposure estimates of the IMF (2024) and ILO (2025) are weighted to clerical and cognitive work; the ILO finds clerical jobs the most exposed of all, and that the highest-risk jobs make up 9.6 per cent of women's employment in high-income countries against 3.5 per cent of men's. For a manufacturer, the parts of the business most affected in the next five years are likely to be quoting, scheduling, purchasing, customer service and quality documentation, not welding. Brynjolfsson, Li and Raymond (2023) found AI assistance helps novices most, which suggests these tools raise the floor rather than remove the roles.

5.4 Implications for UK small and medium-sized manufacturers

For a UK SME the evidence points to four conclusions.

The risk is being out-competed, not automated. The firm-level studies are unambiguous: adopters grow and non-adopters lose share. A 40-person machining business that waits is not protecting its staff; it is exposing them to a competitor's robot.

Automation is an IT project before it is a robot project. A robot cell, a vision system or an AI scheduling tool depends on reliable networking, a segregated operational-technology (OT) network, backed-up programme libraries, patched controllers and identity management for the engineers who maintain them. The UK's low adoption (Department for Business and Trade, 2025) is partly a symptom of weak digital foundations in small firms, and a controller on the same flat network as the office PCs is both a production and a cyber risk. Cyber Essentials expectations on patching and multi-factor authentication extend to any networked equipment in scope, as set out in our post on the April 2026 changes.

Reskill the incumbents. The German finding that automating plants kept their existing people (Dauth et al., 2021) is the most encouraging in the literature, and it depended on workers moving into new tasks inside the firm. That takes training budgets and time, which is what the World Economic Forum (2025) and Make UK (2025a) identify as the gap.

Use the support that now exists. Made Smarter Adoption, the Robotics Adoption Hubs and the innovation funding to 2030 are aimed at firms of this size (Department for Business and Trade, 2025), and the evaluated employment effect of the earlier programme was positive (SQW, 2025).

Our business IT support and cloud services pages describe how we help manufacturers put those foundations in place.

6. Limitations

This is a secondary analysis and inherits the limitations of its sources. The UK density figure of 119 is for 2023 and is cited through a think-tank report rather than the IFR's paid publication; no public 2024 UK density was found. Density is sensitive to the employment denominator, as the China revision shows, so cross-country gaps are orders of magnitude, not precise ratios. The derived output-per-hour series is the author's arithmetic from two ONS indices and will differ slightly from the ONS's own measure; the 2020–2021 values are unusable because of furlough. Workforce-jobs data are June observations. No causal estimate for the UK is attempted, and the country studies end in 2007–2016, before collaborative robots, machine vision and generative AI became cheap, so their findings may not transfer. The IMF, ILO, WEF and McKinsey figures are forecasts and surveys, not outcomes. The paper takes no position on the distributional question of the falling labour share.

7. Conclusion

Between 2015 and 2026 industrial robots spread rapidly in Asia and continental Europe and slowly in the UK. The peer-reviewed record from adopting countries says robots displace specific factory tasks, lower labour's share of value added and shift employment from non-adopting firms to adopters, while leaving total employment roughly unchanged in Europe and reducing it locally in the United States. UK manufacturing has lost about 4 per cent of its jobs since 2015 and nearly 9 per cent since the 2018 peak while producing more with fewer hours, but with a robot density below the world average, which makes robots an unconvincing explanation for the decline. The human-capital story the data support is reallocation under pressure, not decline: know-how moves to the firms that invest in machines and in the people who run them. What is new in 2025–2026 is that automation is arriving through software in the office as much as machines on the floor, and through general-purpose robots whose economics cannot yet be measured. For UK manufacturing SMEs, the answer to "will robots take our jobs?" is that the jobs at risk are at firms that neither automate nor build the IT, skills and security foundations automation requires.

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. https://www.aeaweb.org/articles?id=10.1257/jep.33.2.3 (accessed 25 August 2026).
• Acemoglu, D. and Restrepo, P. (2020) 'Robots and jobs: evidence from US labor markets', Journal of Political Economy, 128(6), pp. 2188–2244. https://www.journals.uchicago.edu/doi/10.1086/705716 (accessed 25 August 2026).
• Acemoglu, D., Lelarge, C. and Restrepo, P. (2020) 'Competing with robots: firm-level evidence from France', AEA Papers and Proceedings, 110, pp. 383–388. https://www.aeaweb.org/articles?id=10.1257/pandp.20201003 (accessed 25 August 2026).
• 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. https://www.aeaweb.org/articles?id=10.1257/jep.29.3.3 (accessed 25 August 2026).
• Autor, D., Chin, C., Salomons, A.M. and Seegmiller, B. (2022) New frontiers: the origins and content of new work, 1940–2018. NBER Working Paper 30389. Cambridge, MA: National Bureau of Economic Research. https://www.nber.org/papers/w30389 (accessed 25 August 2026).
• Brynjolfsson, E., Li, D. and Raymond, L.R. (2023) Generative AI at work. NBER Working Paper 31161. Cambridge, MA: National Bureau of Economic Research. https://www.nber.org/papers/w31161 (accessed 25 August 2026).
• Dauth, W., Findeisen, S., Suedekum, J. and Woessner, N. (2021) 'The adjustment of labor markets to robots', Journal of the European Economic Association, 19(6), pp. 3104–3153. https://sebastian-findeisen.com/papers/adjustmentslaborrobots.pdf (accessed 25 August 2026).
• Department for Business and Trade (2025) Industrial Strategy: Advanced Manufacturing Sector Plan. London: HM Government. https://assets.publishing.service.gov.uk/media/6858622776eec44bf9d71dcf/industrial_strategy_advanced_manufacturing_sector_plan.pdf (accessed 25 August 2026).
• Frey, C.B. and Osborne, M.A. (2013) The future of employment: how susceptible are jobs to computerisation? Oxford Martin School Working Paper. Oxford: University of Oxford. https://www.oxfordmartin.ox.ac.uk/publications/the-future-of-employment (accessed 25 August 2026).
• Graetz, G. and Michaels, G. (2018) 'Robots at work', Review of Economics and Statistics, 100(5), pp. 753–768. https://ideas.repec.org/a/tpr/restat/v100y2018i5p753-768.html (accessed 25 August 2026).
• IFR (2024a) Global robotics race: Korea, Singapore and Germany in the lead. Press release, 10 January 2024. Frankfurt: International Federation of Robotics. https://ifr.org/ifr-press-releases/news/global-robotics-race-korea-singapore-and-germany-in-the-lead (accessed 25 August 2026).
• IFR (2024b) Robot boom in the UK hits all-time high. Press release, 24 September 2024. Frankfurt: International Federation of Robotics. https://ifr.org/downloads/press2018/2024-SEP-24_IFR_press_release_World_Robotics_2024_-_UK.pdf (accessed 25 August 2026).
• IFR (2024c) Global robot density in factories doubled in seven years. Press release, 20 November 2024. Frankfurt: International Federation of Robotics. https://ifr.org/ifr-press-releases/news/global-robot-density-in-factories-doubled-in-seven-years (accessed 25 August 2026).
• IFR (2025a) World Robotics 2025 report: global robot demand in factories doubles over 10 years. Press release, 25 September 2025. Frankfurt: International Federation of Robotics. https://ifr.org/ifr-press-releases/news/global-robot-demand-in-factories-doubles-over-10-years (accessed 25 August 2026).
• IFR (2025b) Humanoid robots: "vision and reality" paper published by IFR. Press release, 14 August 2025. Frankfurt: International Federation of Robotics. https://ifr.org/ifr-press-releases/news/humanoid-robots-vision-and-reality-paper-published-by-ifr (accessed 25 August 2026).
• IFR (2026) Robot density surges in Europe, Asia, and Americas. Press release, 8 April 2026. Frankfurt: International Federation of Robotics. https://ifr.org/ifr-press-releases/news/robot-density-surges-in-europe-asia-and-americas (accessed 25 August 2026).
• ILO (2025) One in four jobs at risk of being transformed by GenAI, new ILO–NASK global index shows. News release, 20 May 2025. Geneva: International Labour Organization. https://www.ilo.org/resource/news/one-four-jobs-risk-being-transformed-genai-new-ilo%E2%80%93nask-global-index-shows (accessed 25 August 2026).
• IMF (2024) Gen-AI: artificial intelligence and the future of work. Staff Discussion Note SDN/2024/001, by Cazzaniga, M. et al. Washington, DC: International Monetary Fund. https://www.imf.org/-/media/files/publications/sdn/2024/english/sdnea2024001.pdf (accessed 25 August 2026).
• Koch, M., Manuylov, I. and Smolka, M. (2021) 'Robots and firms', Economic Journal, 131(638), pp. 2553–2584. https://ideas.repec.org/a/oup/econjl/v131y2021i638p2553-2584..html (accessed 25 August 2026).
• Maier, J. (2017) Made Smarter Review 2017. London: HM Government. https://assets.publishing.service.gov.uk/media/5a74fceced915d502d6cc9dd/20171027_MadeSmarter_FINAL_DIGITAL.pdf (accessed 25 August 2026).
• Make UK (2025a) Making it smarter: global lessons for accelerating automation and digital adoption in UK manufacturing. London: Make UK. https://www.makeuk.org/insights/reports/making-it-smarter-global-lessons-accelerating-automation-digital-adoption-uk (accessed 25 August 2026).
• Make UK (2025b) UK manufacturing: the facts 2025. London: Make UK. https://www.makeuk.org/insights/reports/uk-manufacturing-facts-2025 (accessed 25 August 2026).
• McKinsey Global Institute (2024) A new future of work: the race to deploy AI and raise skills in Europe and beyond. https://www.mckinsey.com/mgi/our-research/a-new-future-of-work-the-race-to-deploy-ai-and-raise-skills-in-europe-and-beyond (accessed 25 August 2026).
• McKinsey Global Institute (2026) Agents, robots, and us: how AI reshapes work and skills in Europe. https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-how-ai-reshapes-work-and-skills-in-europe (accessed 25 August 2026).
• OECD (2023) OECD Employment Outlook 2023: artificial intelligence and the labour market. Paris: OECD Publishing. https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en.html (accessed 25 August 2026).
• ONS (2026a) UK workforce jobs SA: C Manufacturing (thousands), series JWR7, released 18 August 2026. Newport: Office for National Statistics. https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemployeetypes/timeseries/jwr7/lms (accessed 25 August 2026).
• ONS (2026b) IOP: C: Manufacturing: CVMSA, 2023 = 100, series K22A, released 13 August 2026. Newport: Office for National Statistics. https://www.ons.gov.uk/economy/economicoutputandproductivity/output/timeseries/k22a/diop (accessed 25 August 2026).
• ONS (2026c) Productivity hours: Manufacturing C SIC07, 2023 = 100, SA, UK, series DK3V, released 19 May 2026. Newport: Office for National Statistics. https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/labourproductivity/timeseries/dk3v/prdy (accessed 25 August 2026).
• ONS (2026d) AWE: Manufacturing level (£): seasonally adjusted regular pay excluding arrears, series K5DU, released 18 August 2026. Newport: Office for National Statistics. https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/earningsandworkinghours/timeseries/k5du/lms (accessed 25 August 2026).
• ONS (2026e) AWE: Whole economy level (£): seasonally adjusted regular pay excluding arrears, series KAI7, released 18 August 2026. Newport: Office for National Statistics. https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/earningsandworkinghours/timeseries/kai7/lms (accessed 25 August 2026).
• ONS (2026f) Vacancies and jobs in the UK: August 2026. Newport: Office for National Statistics. https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemployeetypes/bulletins/jobsandvacanciesintheuk/august2026 (accessed 25 August 2026).
• SQW (2025) Evaluation of the Made Smarter Innovation Industrial Strategy Challenge Fund: final report. May 2025. https://www.madesmarter.uk/media/gl4jkqoi/evaluation-of-made-smarter-innovation-programme-final-may-2025.pdf (accessed 25 August 2026).
• Tony Blair Institute for Global Change (2024) A new national purpose: the UK's opportunity to lead in next-wave robotics. London: TBI. https://institute.global/insights/tech-and-digitalisation/a-new-national-purpose-the-uks-opportunity-to-lead-in-next-wave-robotics (accessed 25 August 2026).
• World Economic Forum (2025) Future of Jobs Report 2025: 78 million new job opportunities by 2030 but urgent upskilling needed to prepare workforces. Press release, 8 January 2025. Geneva: WEF. https://www.weforum.org/press/2025/01/future-of-jobs-report-2025-78-million-new-job-opportunities-by-2030-but-urgent-upskilling-needed-to-prepare-workforces/ (accessed 25 August 2026).