The Coming Wave and UK Business

AI Transformation Series · Article 2

What the economics of technological transformation tell us about automation, labour, and the opportunities available to UK businesses right now.

Abstract

Artificial intelligence is most commonly discussed as a product category: a collection of tools and applications that improve individual tasks. This article argues that this framing is too narrow. Drawing on the General Purpose Technology framework developed by Timothy Bresnahan and Manuel Trajtenberg, it positions artificial intelligence as a structural economic force comparable in scope to electrification and the internal combustion engine. The article examines the empirical evidence on automation and labour displacement, with particular reference to Carl Benedikt Frey and Michael Osborne’s landmark 2013 study on the susceptibility of occupations to computerisation, and to Daron Acemoglu and Pascual Restrepo’s subsequent research on the task-displacement and task-reinstatement effects of automation. It further draws on Mustafa Suleyman’s analysis of the wave dynamics of transformative technology, Yuval Noah Harari’s examination of information and institutional disruption, and the forecasting frameworks of Kai-Fu Lee and Ray Kurzweil. The implications for UK business owners are examined concretely: which sectors face the greatest displacement risk, where new labour demand is being created, and what positioning choices are available to businesses operating at different scales.

Introduction

Economic history is punctuated by moments when a cluster of technological developments produces not merely incremental improvement but structural transformation. The steam engine did not simply make manufacturing faster; it reorganised the spatial distribution of economic activity, abolished the cottage industry model that had prevailed for centuries, and created entirely new categories of work that had not previously existed. Electrification did not merely replace gas lighting; it enabled the factory floor, the electric motor, and ultimately the entire consumer economy of the twentieth century.

Economists have developed a framework for understanding these episodes: the concept of the General Purpose Technology, a technology characterised by pervasive applicability across multiple sectors of the economy, continuous improvement over time, and the capacity to enable further innovation in the sectors it touches. Timothy Bresnahan and Manuel Trajtenberg, in a 1995 paper that has since been cited more than three thousand times in the academic literature, identified the steam engine, electricity, and information technology as exemplars of this category.

Artificial intelligence meets the definitional criteria of a General Purpose Technology with unusual clarity. It applies across every sector of the economy without exception. It is improving at a rate that has surprised even the researchers most closely involved in its development. And it is already enabling a second wave of innovation: new tools, services, and business models that would not have been conceivable without the underlying capabilities that artificial intelligence provides.

Understanding artificial intelligence as a General Purpose Technology, rather than as a collection of individual products, changes the nature of the strategic conversation for business owners. The question is not simply “which AI tools should I adopt?” It is the broader and more fundamental question: “how does this structural transformation change the competitive landscape in which my business operates, and what positioning choices does that open up or foreclose?”

Theoretical Framework: General Purpose Technologies and Their Dynamics

The Bresnahan-Trajtenberg Framework

Bresnahan and Trajtenberg’s 1995 analysis identified three defining characteristics of General Purpose Technologies. First, pervasiveness: the technology finds application across a wide range of sectors and use cases, rather than being confined to a narrow domain. Second, improvement over time: the technology continues to advance technically after its initial deployment, unlike a specific product that reaches maturity and then stabilises. Third, innovation complementarities: the technology makes it easier to innovate in the sectors it touches, acting as a platform for further technological development.

A fourth characteristic, not explicitly foregrounded by Bresnahan and Trajtenberg but identified by subsequent researchers including Erik Brynjolfsson and Andrew McAfee in their 2014 analysis of the second machine age, is the productivity paradox. General Purpose Technologies consistently produce a period of measured productivity stagnation before their economic benefits become visible in aggregate statistics. This occurs because the organisational changes required to realise the technology’s full potential, the redesign of processes, the retraining of workers, the restructuring of business models, take time and impose costs before they deliver returns.

Brynjolfsson and McAfee estimated that the information technology revolution produced a productivity paradox lasting approximately a decade and a half before the gains became visible at the macroeconomic level. If a similar lag applies to artificial intelligence, the period of apparent disruption and uncertainty currently being experienced may be precisely the period immediately preceding the most significant productivity gains.

Wave Dynamics: Suleyman’s Analysis

Mustafa Suleyman, in his 2023 book The Coming Wave, offers a complementary analytical framework focused on the dynamics of technological diffusion rather than the structural economics of productivity. Suleyman argues that transformative technologies share a common pattern: an initial period of contained development within research environments, followed by a rapid release into the broader economy as the technology crosses cost and capability thresholds, succeeded by a wave of adoption that reshapes existing institutions faster than those institutions can adapt to accommodate it.

What distinguishes the current wave of artificial intelligence, in Suleyman’s analysis, is the combination of breadth and pace. Earlier transformative technologies were broad but slow-moving: electrification took decades to penetrate the full range of economic activity. Other technologies were rapid but narrow in scope. Artificial intelligence, he argues, is simultaneously broad in its applicability and rapid in its improvement, a combination that makes the current transition unusual even by historical standards.

Yuval Noah Harari, approaching the same phenomenon from the perspective of information theory and institutional history in his 2023 book Nexus, adds a further dimension. Harari argues that artificial intelligence differs from previous General Purpose Technologies in one crucial respect: it is the first technology capable of generating, processing, and distributing information autonomously, without requiring human intermediation at each step. This, he contends, represents not merely a quantitative improvement in information processing but a qualitative shift in the relationship between human institutions and the information systems they depend upon.

Core Analysis: The Evidence on Automation and Labour

Frey and Osborne: The 47 Percent Question

In 2013, Carl Benedikt Frey and Michael Osborne of the Oxford Martin School published a working paper entitled “The Future of Employment: How Susceptible Are Jobs to Computerisation?” The paper has since been cited more than eight thousand times in the academic literature, making it one of the most influential empirical studies in recent economic research.

Frey and Osborne analysed 702 occupational categories from the United States Bureau of Labor Statistics O*NET database, assessing the susceptibility of each occupation to computerisation based on its reliance on three categories of tasks identified as particularly resistant to automation in 2013: perception and manipulation (fine motor tasks in unpredictable environments), creative intelligence (generating novel ideas or artefacts), and social intelligence (negotiating, persuading, caring for others). Their central finding was that 47 percent of total US employment was at high risk of automation within the following ten to twenty years.

The Frey and Osborne figure has generated substantial subsequent debate. Some researchers have argued that it overstates automation risk by focusing on occupational categories rather than the specific tasks within those occupations, many of which may not be automatable even if the occupation as a whole appears susceptible. Others have argued that it understates the risk by not accounting for the accelerating pace of capability development in artificial intelligence systems after 2013.

What is not in dispute is the directional finding: a substantial proportion of existing occupational tasks, concentrated particularly in routine cognitive work and structured manual work, face meaningful displacement risk from the current and foreseeable generation of artificial intelligence capabilities. For UK businesses, many of which rely heavily on precisely these categories of work in functions such as bookkeeping, data entry, document processing, scheduling, and customer service, this finding is directly relevant to workforce planning and competitive strategy.

Acemoglu and Restrepo: Displacement and Reinstatement

Daron Acemoglu and Pascual Restrepo have produced a series of highly cited studies that add important nuance to the automation debate. Their 2018 paper “The Race Between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment,” published in the American Economic Review, and their 2019 Journal of Economic Perspectives paper “Automation and New Tasks: How Technology Displaces and Reinstates Labor,” develop a theoretical framework that distinguishes between the displacement effect of automation, where machines substitute for human labour in existing tasks, and the reinstatement effect, where new tasks are created that are complementary to automation and require human labour.

The central insight of their framework is that the net effect of automation on employment and wages depends on the balance between displacement and reinstatement. When automation is primarily displacing existing tasks without creating sufficient new tasks to absorb the displaced labour, the effect on aggregate employment and wages is negative. When reinstatement is strong enough, the net effect can be positive even as individual occupations are disrupted.

Their empirical research, examining US labour market data from 1980 to 2016, found that the displacement effect of automation has been stronger than the reinstatement effect during this period, contributing to labour market polarisation: the growth of high-skill, high-wage employment and low-skill, low-wage employment, with contraction in the middle-skill, middle-wage occupations most susceptible to routine cognitive automation. Their 2020 paper in the Journal of Political Economy, “Robots and Jobs: Evidence from US Labor Markets,” found that each additional industrial robot per one thousand workers reduced employment by 0.2 percent and wages by 0.42 percent in the affected area.

Kai-Fu Lee and the Comparative Advantage of Human Judgment

Kai-Fu Lee, in his 2018 book AI Superpowers: China, Silicon Valley, and the New World Order and his co-authored 2021 work AI 2041: Ten Visions for Our Future, offers a framework for thinking about where human comparative advantage persists in an economy shaped by artificial intelligence. Lee distinguishes between tasks characterised by optimisation within a well-defined problem space, which are susceptible to automation, and tasks characterised by the exercise of judgment under uncertainty, creative synthesis, and interpersonal relationship, which remain domains of human advantage.

For UK business owners, this framework suggests a strategic question: which of the tasks currently performed within the business fall into the first category, and which into the second? The former are candidates for automation, with associated reductions in cost and potential improvements in consistency. The latter represent the dimensions of competitive differentiation that artificial intelligence is least likely to replicate in the near term.

Ray Kurzweil’s forecasting framework, presented in his 2024 book The Singularity is Nearer, offers a more optimistic projection of this dynamic. Kurzweil argues that the pattern of exponential improvement in artificial intelligence capabilities, combined with exponential improvement in the biological and cognitive augmentation of human intelligence, will produce not a displacement of human economic agency but a radical expansion of it. New domains of valuable activity will emerge faster than existing domains are automated.

The debate between the displacement-oriented reading of Acemoglu and Restrepo and the reinstatement-optimistic reading of Kurzweil is not resolved in the empirical literature. What is clear is that the outcome depends significantly on choices made by businesses, governments, and individuals during the current transitional period. The businesses that position themselves on the right side of the displacement-reinstatement balance, by identifying which of their activities are genuinely automatable and investing the liberated capacity in tasks requiring human judgment, are most likely to benefit from the transition rather than being harmed by it.

The strategic frame for UK business: The question is not whether artificial intelligence will affect your business. The economic evidence is unambiguous that it will. The question is whether your business is positioned to capture the reinstatement effects, the new forms of value creation enabled by automation, or whether it will experience only the displacement effects, the loss of existing revenue streams and workforce capacity without replacement.

Implications for UK Business Owners

Identify the Automatable Core

The Frey-Osborne and Acemoglu-Restrepo frameworks provide a useful diagnostic lens. Within any business, some proportion of activity consists of structured, rule-governed, repetitive tasks: tasks that require accuracy and consistency rather than judgment and creativity. These activities are the candidates for near-term automation. Identifying them explicitly, rather than waiting for displacement to occur reactively, allows the business to control the transition: to choose the timing, manage the workforce implications proactively, and reinvest the liberated capacity into higher-value activities.

Invest in the Human Complement

Lee’s framework of human comparative advantage suggests that the activities most resistant to automation are those that involve judgment under uncertainty, genuine interpersonal relationship, ethical navigation, and creative synthesis. For most UK businesses, this encompasses client relationship management, strategic decision-making, complex problem-solving, and the domain expertise that is built through years of professional practice.

The productivity paradox identified by Brynjolfsson and McAfee suggests that realising the value of artificial intelligence requires not merely adopting the tools but reorganising the business around them: redesigning processes, redefining roles, and investing in the human capabilities that complement rather than duplicate machine performance. This reorganisation takes time and imposes transitional costs. The businesses that begin it earlier, during the current period of widespread uncertainty, will be better positioned when the productivity gains become visible.

The UK Context

The United Kingdom faces the challenge of artificial intelligence adoption within a specific macroeconomic context characterised by relatively low productivity growth compared to other comparable economies, a service-dominated economy in which many of the most automatable tasks are concentrated, and a skills infrastructure that is in the process of adaptation but has not yet been restructured to reflect the requirements of an AI-integrated economy.

For business owners operating in this context, the strategic priority is clarity about which of the forces described in this article, displacement or reinstatement, is most likely to affect their specific sector and business model, and what proactive positioning choices are available to shift that balance toward the latter. That clarity is not available from social media commentary or general technology journalism. It requires structured analysis of the specific commercial environment in which each business operates.

Conclusion

The General Purpose Technology framework developed by Bresnahan and Trajtenberg provides a more useful lens for understanding the current moment of artificial intelligence development than the product-category framing that dominates popular commentary. Artificial intelligence is not a set of tools to be adopted or rejected. It is a structural transformation of the economic environment in which all businesses operate, comparable in its eventual scope to electrification and information technology.

The empirical evidence of Frey and Osborne, and of Acemoglu and Restrepo, indicates that this transformation will produce substantial displacement of existing occupational tasks, concentrated in routine cognitive and structured manual work. It also indicates, consistent with the historical pattern of General Purpose Technology adoption, that new tasks and new forms of value creation will emerge, though the pace and distribution of that reinstatement is not guaranteed and depends on choices made by businesses and policymakers during the transitional period.

For UK business owners, the relevant question is not whether this transformation will occur, but how to position a specific business to navigate it with strategic intentionality rather than reactive adaptation. That positioning begins with the kind of clear analytical framework that this article has attempted to provide, and it continues with the specific decisions about automation, investment, and workforce development that only the business owner is positioned to make.

Work With Zazentax

Zazentax supports business owners with the financial clarity and strategic analysis needed to navigate transformation. Whether you are restructuring operations, planning for growth, or managing the tax implications of business change, our team brings both technical expertise and commercial understanding to the work.

Visit www.zazentax.com to learn more or to arrange a consultation.

This is Article 2 of the Zazentax AI Transformation Series. Article 1 examined the neuroscience of adaptation. Article 3 examines how the falling cost of prediction changes the nature of business judgment and where human expertise remains irreplaceable.

References

1. Bresnahan, T. F., & Trajtenberg, M. (1995). General purpose technologies: “Engines of growth”? Journal of Econometrics, 65(1), 83-108.
2. Brynjolfsson, E., & McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W. W. Norton & Company.
3. Frey, C. B., & Osborne, M. A. (2013). The future of employment: How susceptible are jobs to computerisation? Oxford Martin School Working Paper.
4. Acemoglu, D., & Restrepo, P. (2018). The race between man and machine: Implications of technology for growth, factor shares, and employment. American Economic Review, 108(6), 1488-1542.
5. Acemoglu, D., & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3-30.
6. Acemoglu, D., & Restrepo, P. (2020). Robots and jobs: Evidence from US labor markets. Journal of Political Economy, 128(6), 2188-2244.
7. Suleyman, M. (2023). The Coming Wave: Technology, Power, and the Twenty-First Century’s Greatest Dilemma. Crown Publishers.
8. Harari, Y. N. (2023). Nexus: A Brief History of Information Networks from the Stone Age to AI. Signal / McClelland & Stewart.
9. Lee, K.-F. (2018). AI Superpowers: China, Silicon Valley, and the New World Order. Houghton Mifflin Harcourt.
10. Lee, K.-F., & Qiufan, C. (2021). AI 2041: Ten Visions for Our Future. Currency.
11. Kurzweil, R. (2024). The Singularity is Nearer. Viking.