University of Kwazulu-Natal

Working Paper 25 021 Generative Ai And The Nature Of Work

Bonisiwe Shabane
-
working paper 25 021 generative ai and the nature of work

Recent advances in artificial intelligence (AI) technology demonstrate considerable potential to complement human capital intensive activities. While an emerging literature documents wide-ranging productivity effects of AI, relatively little attention has been paid to how AI might change the nature of work itself. How do individuals, especially those in the knowledge economy, adjust how they work when they start using AI? Using the setting of open source software, we study individual level effects that AI has on task allocation. We exploit a natural experiment arising from the deployment of GitHub Copilot, a generative AI code completion tool for software developers. Leveraging millions of work activities over a two year period, we use a program eligibility threshold to investigate the impact of AI technology on the task allocation of software developers within a quasi-experimental regression...

We find that having access to Copilot induces such individuals to shift task allocation towards their core work of coding activities and away from non-core project management activities. We identify two underlying mechanisms driving this shift - an increase in autonomous rather than collaborative work, and an increase in exploration activities rather than exploitation. The main effects are greater for individuals with relatively lower ability. Overall, our estimates point towards a large potential for AI to transform work processes and to potentially flatten organizational hierarchies in the knowledge economy. You are seeing this because EconStor has set up <a href="https://github.com/TecharoHQ/anubis">Anubis</a> to protect the server against the scourge of AI companies aggressively scraping websites. This can and does cause downtime for the websites, which makes their resources inaccessible for everyone.

Anubis is a compromise. Anubis uses a Proof-of-Work scheme in the vein of Hashcash, a proposed proof-of-work scheme for reducing email spam. The idea is that at individual scales the additional load is ignorable, but at mass scraper levels it adds up and makes scraping much more expensive. Ultimately, this is a placeholder solution so that more time can be spent on fingerprinting and identifying headless browsers (EG: via how they do font rendering) so that the challenge proof of work page... Please note that Anubis requires the use of modern JavaScript features that plugins like JShelter will disable. Please disable JShelter or other such plugins for this domain.

Protected by Anubis From Techaro. Made with ❤️ in 🇨🇦. This Harvard-Microsoft-GitHub collaboration examines how generative AI changes how people work—beyond just how fast. Instead of only focusing on overall productivity, the authors look at shifts in task allocation caused by GitHub Copilot, an AI code assistant trained on LLMs. The research uses a quasi-experimental regression discontinuity design to evaluate the behavior of over 180,000 developers on GitHub from 2022 to 2024. A subset of developers—ranked just above an internal threshold—received free access to Copilot, while those just below did not.

This setup provides a credible causal estimate of AI’s effects on labor allocation. The authors propose a theoretical model of work adjustment under AI assistance. The model’s predictions match the empirical outcomes: generative AI reduces coordination costs and reallocates effort toward more valued, technical contributions. Generative AI is poised to reshape knowledge work. By studying GitHub Copilot’s impact on open-source developers, this paper shows AI access leads to a reallocation of effort—less time on coordination, more on core tasks. These changes suggest AI may reduce collaboration burdens and flatten hierarchies in digital labor markets.

This is one of the first large-scale causal studies showing that generative AI doesn’t just speed up tasks—it rebalances them. It reveals how AI shifts cognitive work away from overhead and toward creative, high-value contributions. For governance, this opens new questions: Are these shifts good for innovation? Fair for teams? Sustainable for collaboration? I am a postdoctoral fellow at the Laboratory for Innovation Science at Harvard.

I am an applied microeconomist with research interests at the intersection of digital economics, labor and productivity, industrial organization, and socio-technical networks. Specifically, my work has centered around the private provision of public goods, productivity in open collaboration, and welfare effects within the context of open source software (OSS) ecosystems. Recent advances in artificial intelligence (AI) technology demonstrate considerable potential to complement human capital intensive activities. While an emerging literature documents wide-ranging productivity effects of AI, relatively little attention has been paid to how AI might change the nature of work itself. How do individuals, especially those in the knowledge economy, adjust how they work when they start using AI? Using the setting of open source software, we study individual level effects that AI has on task allocation.

We exploit a natural experiment arising from the deployment of GitHub Copilot, a generative AI code completion tool for software developers. Leveraging millions of work activities over a two year period, we use a program eligibility threshold to investigate the impact of AI technology on the task allocation of software developers within a quasi-experimental regression... We find that having access to Copilot induces such individuals to shift task allocation towards their core work of coding activities and away from non-core project management activities. We identify two underlying mechanisms driving this shift - an increase in autonomous rather than collaborative work, and an increase in exploration activities rather than exploitation. The main effects are greater for individuals with relatively lower ability. Overall, our estimates point towards a large potential for AI to transform work processes and to potentially flatten organizational hierarchies in the knowledge economy.

Innovation, a key driver of competitive advantage, increasingly occurs through informal partnerships between firms collaborating on public goods. However, competitive forces may influence contribution patterns. Existing studies primarily consider the role competition plays in closed and private innovation. Therefore, we utilize data on millions of firm contributions to open source software, a crowdsourced public good, to measure the relationship between the level of competition a firm faces and its participation in informal... We further create a novel measure of the labor competition a firm faces using Lightcast data on job postings. While the relationship between labor and product market competition is nuanced and warrants further investigation, labor competition is particularly important in innovative and technology-focused contexts.

We find that labor market competition and open source contributions exhibit an inverted U-relationship. The current findings hint at open innovation being a potentially strategic choice in hiring decisions during times of low labor market power of firms. We empirically examine the extent to which peer effects influence the private provision of public goods. In the case of public information goods, peer contribution may facilitate or otherwise incentivize further contribution from others, effectively subsidizing private provision. Using the setting of Open Source Software (OSS) contribution, we first utilize a reduced form approach to derive causal estimates of net peer effects in public goods contribution by exploiting a peers-of-peers identification strategy. We next develop a structural model of peer-influenced public good provision that both (1) separates extensive and intensive margin contribution decisions and (2) decomposes contribution into marginal private benefits and costs.

We apply these methodologies using a sample of peer contribution histories for 2,287 OSS projects hosted on the GitHub collaboration platform. Both reduced form and structural approaches suggest peer effects are much stronger along the extensive margin than the intensive margin. Contemporaneous intensive margin effects, while heterogenous across time and projects, are small and centered around zero, suggesting that strategic complementarity and substitution in peer contribution likely offset one another. Our counterfactual analysis suggests (extensive margin) peer effects account for nearly 56% of cumulative aggregate contribution in the sample, which translates to a value-added of 1-1.5 million software developer labor hours. These results support the notion that OSS is largely developed by disproportionate efforts from smaller groups of dedicated core maintainers, who integrate incremental contributions from the wider community, and casts doubt on the promise... Developers of software projects can leverage the functionality of existing open source projects.

This practice can potentially lower the cost of development albeit at the inherent risk of relying on external components. A “downstream” project maintainer can choose to “import” elements of an “upstream” project to outsource functionality, but is uncertain how future changes in this dependency project may expose her own project to software faults... Software dependency networks therefore represent a “digital supply chain”, an ecosystem of interdependent public goods that confer an intricate set of both positive and negative externalities for project maintainers and end users. Focusing on microeconomic fundamentals of the dependency management problem faced by the risk averse project maintainer, we use both reduced form and structural approaches to study how dependency networks create value, what forces shape... We use a sample of open source software projects from the Node.js JavaScript packaging ecosystem for which contribution and dependency formation decisions are observed in real-time. Finally, we consider several policy interventions that can improve equilibrium welfare.

In particular, we find that removing less that 1% of core projects can reduce aggregate project quality by more than 5% for the remaining peers.

People Also Search

Recent Advances In Artificial Intelligence (AI) Technology Demonstrate Considerable Potential

Recent advances in artificial intelligence (AI) technology demonstrate considerable potential to complement human capital intensive activities. While an emerging literature documents wide-ranging productivity effects of AI, relatively little attention has been paid to how AI might change the nature of work itself. How do individuals, especially those in the knowledge economy, adjust how they work ...

We Find That Having Access To Copilot Induces Such Individuals

We find that having access to Copilot induces such individuals to shift task allocation towards their core work of coding activities and away from non-core project management activities. We identify two underlying mechanisms driving this shift - an increase in autonomous rather than collaborative work, and an increase in exploration activities rather than exploitation. The main effects are greater...

Anubis Is A Compromise. Anubis Uses A Proof-of-Work Scheme In

Anubis is a compromise. Anubis uses a Proof-of-Work scheme in the vein of Hashcash, a proposed proof-of-work scheme for reducing email spam. The idea is that at individual scales the additional load is ignorable, but at mass scraper levels it adds up and makes scraping much more expensive. Ultimately, this is a placeholder solution so that more time can be spent on fingerprinting and identifying h...

Protected By Anubis From Techaro. Made With ❤️ In 🇨🇦.

Protected by Anubis From Techaro. Made with ❤️ in 🇨🇦. This Harvard-Microsoft-GitHub collaboration examines how generative AI changes how people work—beyond just how fast. Instead of only focusing on overall productivity, the authors look at shifts in task allocation caused by GitHub Copilot, an AI code assistant trained on LLMs. The research uses a quasi-experimental regression discontinuity des...

This Setup Provides A Credible Causal Estimate Of AI’s Effects

This setup provides a credible causal estimate of AI’s effects on labor allocation. The authors propose a theoretical model of work adjustment under AI assistance. The model’s predictions match the empirical outcomes: generative AI reduces coordination costs and reallocates effort toward more valued, technical contributions. Generative AI is poised to reshape knowledge work. By studying GitHub Cop...