Magnus Lodefalk

This paper investigates the economic and societal impacts of Artificial Intelligence (AI) in the public sector, focusing on its potential to enhance productivity and mitigate labour shortages. Employing detailed administrative data and novel occupational exposure measures, we simulate future scenarios over a 20-year horizon, using Sweden as an illustrative case. Our findings indicate that advances in AI development and uptake could significantly alleviate projected labour shortages and enhance productivity. However, outcomes vary substantially across sectors and organisational types, driven by differing workforce compositions. Complementing the economic analysis, we identify key challenges that hinder AI’s effective deployment, including technical limitations, organisational barriers, regulatory ambiguity, and ethical risks such as algorithmic bias and lack of transparency. Drawing from an interdisciplinary conceptual framework, we argue that AI’s integration in the public sector must address these socio-technical and institutional factors comprehensively. To unlock AI’s full potential, substantial investments in technological infrastructure, human capital development, regulatory clarity, and robust governance mechanisms are essential. Our study thus contributes both novel economic evidence and an integrated societal perspective, informing strategies for sustainable and equitable public-sector digitalisation

This paper documents novel facts on within-occupation task and skill changes over the past two decades in Germany. In a second step, it reveals a distinct relationship between occupational work content and exposure to artificial intelligence (AI) and automation (robots). Workers in occupations with high AI exposure, perform different activities and face different skill requirements, compared to workers in occupations exposed to robots. In a third step, the study uses individual labour market biographies to investigate the impact on wages between 2010 and 2017. Results indicate a wage growth premium in occupations more exposed to AI, contrasting with a wage growth discount in occupations exposed to robots. Finally, the study further explores the dynamic influence of AI exposure on individual wages over time, uncovering positive associations with wages, with nuanced variations across occupational groups

This paper investigates the impact of artificial intelligence (AI) on hiring and employment, using the universe of job postings published by the Swedish Public Employment Service from 2014-2022 and universal register data for Sweden. We construct a detailed measure of AI exposure according to occupational content and find that establishments exposed to AI are more likely to hire AI workers. Survey data further indicate that AI exposure aligns with greater use of AI services. Importantly, rather than displacing non-AI workers, AI exposure is positively associated with increased hiring for both AI and non-AI roles. In the absence of substantial productivity gains that might account for this increase, we interpret the positive link between AI exposure and non-AI hiring as evidence that establishments are using AI to augment existing roles and expand task capabilities, rather than to replace non-AI workers

We use individual survey data providing detailed information on stress, technology adoption, and work, worker, and employer characteristics, in combination with recent measures of AI and robot exposure, to investigate how new technologies affect worker stress. We find a persistent negative relationship, suggesting that AI and robots could reduce the stress level of workers. We furthermore provide evidence on potential mechanisms to explain our findings. Overall, the results provide suggestive evidence of modern technologies changing the way we perform our work in a way that reduces stress and work pressure

We unbox developments in artificial intelligence (AI) to estimate how exposure to these developments affect firm-level labour demand, using detailed register data from Denmark, Portugal and Sweden over two decades. Based on data on AI capabilities and occupational work content, We develop and validate a time-variant measure for occupational exposure to AI across subdomains of AI, including language modelling. According to our model, white collar occupations are most exposed to AI, and especially white collar work that entails relatively little social interaction. We illustrate its usefulness by applying it to near-universal data on firms and individuals from Sweden, Denmark, and Portugal, and estimating firm labour demand regressions. We find a positive (negative) association between AI exposure and labour demand for highskilled white (blue) collar work. Overall, there is an up-skilling effect, with the share of white-collar to blue collar workers increasing with AI exposure. Exposure to AI within the subdomains of image and language are positively (negatively) linked to demand for high-skilled white collar (blue collar) work, whereas other AI-areas are heterogeneously linked to groups of workers

Artificial intelligence (AI) is expected to reshape labor markets, yet causal evidence remains scarce. We exploit a novel Swedish subsidy program that encouraged small and mid-sized firms to adopt AI. Using a synthetic difference-in-differences design comparing awarded and non-awarded firms, we find that AI subsidies led to a sustained increase in job postings over five years, but with no statistically detectable change in employment. This pattern reflects hiring signals concentrated in AI occupations and white-collar roles. Our findings align with task-based models of automation, in which AI adoption reconfigures work and spurs demand for new skills, but hiring frictions and the need for complementary investments delay workforce expansion

We show that the age composition of employment within Swedish employers shifts after the arrival of generative AI, with no corresponding reduction in aggregate labour demand. Using 4.6 million job advertisements from Sweden’s largest recruitment platform, we find that the broad decline in postings since 2022 aligns with monetary tightening rather than AI, exploiting Sweden’s seven-month gap between the Riksbank’s first rate hike and the launch of ChatGPT as a timing test. We then use full-population employer– employee register data and an employer-level difference-in-differences design to estimate how AI exposure affects employment composition across six age groups. An event study documents an accelerating decline in employment of 22–25-year-olds in high-AI-exposure occupations, reaching 5.5 per cent by early 2025 relative to less exposed occupations within the same employers, while employment of workers over 50 rose by 1.3 per cent. The widening age gradient suggests that generative AI reshapes hiring composition rather than aggregate demand, with the adjustment burden falling disproportionately on entry-level workers

Using two waves of nationally representative Danish firm surveys linked to employer–employee administrative registers, we study how adoption varies across artificial intelligence (AI) and related advanced technologies. We show that AI adoption is highly technology-specific. While firm size and digital infrastructure predict adoption broadly, workforce composition operates through distinct channels: STEM-educated workforces predict core AI adoption, whereas non-STEM university-educated workforces are associated with generative AI adoption, indicating different human capital complementarities. The factors associated with adoption differ from those predicting deployment breadth: firm size and digital maturity matter for both, whereas workforce composition primarily predicts adoption alone. Machine learning and natural language processing are deployed across multiple business functions, whereas other advanced technologies remain concentrated in specific operational domains. Individual-level evidence provides a foundation for these patterns, with awareness of workplace AI usage concentrated among managers and high-skilled workers. Self-reported AI knowledge is higher among younger and more educated individuals. Finally, commonly used occupational AI exposure measures vary substantially in their ability to predict observed adoption, with benchmark-based measures outperforming patent-based and LLM-focused alternatives. These findings show that treating AI as a monolithic category obscures economically meaningful variation in who adopts, what they deploy, and how well existing measures capture it.