From AI adoption to an ageing workforce, the demand for skills is changing fast. This session examined what firms need, where the current system is falling short, and how policy can better support workers and employers through the transition.
The session connected productivity growth to the movement and development of people, skills and knowledge. The presentations showed that productivity gains may come from improving the proficiency of existing workers, widening access to specialised talent, and enabling knowledge to travel across occupational, organisational and national boundaries. At the same time, the discussion stressed that these gains are not automatic. Skills interventions have costs, AI can redistribute control and professional status, language changes can create cultural and labour-market trade-offs, and migration effects vary across industries and regions. A consistent message was that policy needs to consider both aggregate productivity gains and how benefits, costs and decision-making power are distributed.
The first paper investigated how AI changes role relations and professional status in collaborative work. Using interviews and archival material from the games sector, the study focused on the close interaction of artists, designers and programmers during pre-production and full production. It contrasted broad technologies, where workers retain discretion over whether and how to use AI, with specific technologies imposed through client contracts or licensing relationships.
The researchers identified four forms of bypassing practice. Overlapping roles occurred when occupational groups used AI to move into one another’s tasks without major conflict. Excluding roles arose when, for example, a lead designer created a pitch without the usual input from artists or programmers. Fragmenting roles appeared in buyer-supplier relationships, where the buyer used AI-enabled processes while retaining more control over quality decisions. Transferring roles occurred when licensed AI tools automated activity within a games studio while related technical and creative work reappeared inside specialist AI firms. The paper argued that analysis should look beyond whether AI automates or augments an individual job and instead track effects across collaborative teams, organisational boundaries and value chains. Discussion covered consumer backlash against AI-generated content, worker autonomy, rotation out of subcontracted work and the possible link between reduced professional status and productivity.
The second paper used universities as a setting to study whether switching the language of work to English can widen labour-market matching and attract stronger international talent. Degree programmes provide a clear point at which teaching officially changes language, allowing the researchers to compare hiring and publication outcomes before and after the switch. The analysis covered eight Dutch universities and used universities ranked above the highest-ranked Dutch institution as a never-treated comparison group. Outcomes included the number of new hires, the international composition of hiring and publication-based measures of research ability.
The presentation reported large post-switch changes. Total publication-based ability among new hires increased by 162 per cent over the baseline, the number of new hires increased by 113 per cent, and foreign hiring increased by 251 per cent. Average publication-based ability among new hires rose by 69 per cent. The speaker stressed that the expansion of English-language programmes also made universities much larger, helping to explain the scale of the aggregate effect. The discussion considered institutional capacity, differences between universities, potential changes in evaluation criteria, student and local labour-market outcomes, and the trade-off between research productivity and the protection of local language and culture.
The third paper examined whether skilled migration raises AI-specific innovation in the regional industries that employ migrant workers. The study combined AI-classified patent data with United Kingdom household survey data and ONS controls for 2011 to 2022. Migrant workers were defined as people born outside the United Kingdom and aged 16 to 64, regardless of citizenship. The analysis was conducted at industry-region level and used a shift-share instrumental-variable approach based on the earlier geographic distribution of migrants to address possible self-selection into stronger local economies.
The results presented indicated that an additional 1,000 migrant workers in an industry-region cell was associated with a two-percentage-point increase in the AI share of patenting. When the analysis focused on highly skilled migrants, particularly those in STEM and research roles, the reported effect rose to six percentage points. The findings were described as robust to checks excluding London, examining non-AI patenting and comparing periods before and after the United Kingdom’s exit from the European Union. Discussion suggested weighting patents by citations to capture quality, clarifying how AI patents are distributed across industrial strategy sectors, and recognising international students and skilled migrants as potential contributors to innovation and productivity.