Beyond job displacement: Modelling the economic impact of AI and automation

Much of the emerging literature on AI uses task-based analysis to estimate which activities could be automated and then considers the potential implications for employment. While valuable, this can provide only a partial picture of how economies respond to technological change.

Cambridge Econometrics developed a new modelling framework for DSIT that embedded this task-based analysis within the broader economics of innovation. Rather than assuming that automating tasks translates directly into job losses, the model examines how businesses reorganise production in response to each wave of automation, and how these changes affect innovation, productivity, demand, occupations, skills and employment.

Approach

We developed an interactive economic model linking task-level automation to business adaptation, innovation, occupational change and wider economic outcomes.

The starting point is the changing technological capability of AI and the tasks that could consequently be automated. But this is only the initial shock. As tasks become automatable, the relative costs and capabilities of different activities change. Businesses respond by reorganising production, changing how automated and human activities are combined and altering their demand for activities that complement or substitute for the new technology.

These adaptations can generate both process and product innovation. Process innovation allows existing products and services to be produced more efficiently. Product innovation can improve existing products or create entirely new goods, services and activities.

The economic effects then depend on how markets respond. Productivity improvements can reduce costs and prices, generating additional demand for existing products, while product innovation can create new sources of demand. The scale of these responses depends on factors including the elasticities of supply and demand.

The model also captures how work itself changes. Occupations are bundles of activities, some of which can be automated while others become more valuable because they complement the technology. As businesses reorganise production, the nature of jobs and the tasks undertaken by workers may change. This can lead to new skills requirements, opportunities for upskilling or deskilling as well as changes in the number of jobs.

An interactive Excel model and dashboard allowed DSIT analysts to vary key assumptions and explore alternative scenarios rather than relying on a single forecast.

Key insights

The modelling highlighted three important insights.

Task automation does not translate mechanically into job displacement. Automation changes the costs and capabilities available to businesses. Firms then reorganise production in response, meaning that some activities contract while complementary activities can expand. The eventual impact on employment depends on this wider process of adaptation.

The balance between productivity and demand is critical. Process innovation can reduce the labour required to produce a given level of output, but the resulting productivity gains can lower costs and prices and stimulate additional demand. Product innovation can create further demand by enabling new and improved goods and services. Employment outcomes therefore depend on how supply and demand respond to technological change, not simply on how much labour can technically be automated.

AI is likely to change the content of jobs as well as their number. As firms reorganise production, occupations are rebundled around a changing mix of automated, complementary and substitutable activities. This can change skill requirements within occupations, producing upskilling or deskilling even where the occupation itself remains.

The model therefore moves beyond the question of which tasks AI can automate to examine how businesses, workers and markets adapt once those new capabilities become available.

Impact

The work gave DSIT a framework for understanding AI as a process of innovation and economic transformation rather than simply a source of job displacement.

By connecting the emerging task-based AI literature with established economic thinking on innovation, productivity, supply and demand and occupational change, the model provided a structured way to explore how different assumptions about AI could translate into different outcomes for output, productivity and employment.

Rather than producing a single prediction about the future of AI, the interactive model allowed analysts to explore alternative technological and economic scenarios and understand the mechanisms driving the results.

The modelling was developed collaboratively with DSIT and subsequently formed part of its analytical work on the economic and labour-market implications of AI. Its potential relationship with wider macroeconomic and fiscal analysis was also explored with HM Treasury.

Outcomes

The project provided government with a richer framework for assessing the economic consequences of AI.

Its central contribution was to distinguish technological capability from economic outcome. Knowing which tasks AI can automate is an important starting point, but the ultimate consequences depend on how firms reorganise production; how automation enables process and product innovation; how supply and demand respond; how occupations are rebundled around complementary and substitutable activities; and how workers' skills adapt.

By bringing these mechanisms together, Cambridge Econometrics helped shift the analysis from a narrow question about jobs at risk from automation towards a broader understanding of how AI could transform production, innovation and work across the UK economy.