AI and the reorganization of RSE work

Abstract

Coding agents like Claude Code and GitHub Copilot have significant implications for the way software is built and maintained in the sciences. Some of these implications center around changes in practice, such as greater or worse productivity and the need for verification practices (He et al., 2025; O'Brien, 2025a; O’Brien, 2025b; Besser et al., 2026; O'Brien and Eisty, 2026). Another dimension of concern, however, is how research software development work will be reorganized around the presence of coding agents. Broader discourse has focused on the possibility of replacement or simple automation of software engineering work by agents, but, as with prior developments, the change is more likely to involve a more complex re-ordering of the work of producing software (Jarrahi, 2018; see also Gesing, 2025). Emerging work in industrial and open source contexts, for instance, suggests that coding agents can affect the interactions around code review (Song et al., 2024; Agarwal et al., 2026a), assume different roles in the development process (Agarwal et al., 2026b), and can change where particular kinds of development work fall in a team or community (Xu et al., 2025). This poster explores these shifts in the context of RSE work by examining dynamics of re-delegation going on around the adoption of coding agents by researchers and RSEs. Based on interviews with 25 RSEs and RSE managers in the US, we first outline these RSEs’ rationales about what makes research software development either amenable or unsuitable for the use of AI coding agents. These include niche codebases, siloed working environments, savvy end users, prevalence of prototyping, and the importance of problem formation. We then outline three dynamics of re-delegation, each of which represents a shift in the way work is divided up and assigned (or taken on) amongst human and non-human actors (Ribes et al., 2013). These are: enabling end-user development, delegating implementation, and taking on AI training roles. In some cases our interlocutors report quite different and sometimes contradictory experiences across these categories. This raises two questions that can help stimulate further discussion on this topic: 1) What aspects of RSE working environments might contribute to RSEs’ different experiences with AI coding agents? 2) What organizational changes could RSEs or RSE teams make that might mitigate some of the problems associated with the growing usage of AI coding agents while making best use of those agents’ benefits? References Agarwal, S., Miller, C., Kästner, C., & Vasilescu, B. (2026). 3100 opinions on code review in an AI world: building causal theory from practitioner discourse. arXiv preprint arXiv:2607.07980. Agarwal, S., He, H., & Vasilescu, B. (2026). AI IDEs or autonomous agents? measuring the impact of coding agents on software development. arXiv preprint arXiv:2601.13597. Besser, S. A., Jensen, E. A., & Katz, D. S. (2026). How generative AI is shaping research software development and maintenance at a research-intensive university. Open Research Europe, 6, 56. Gesing, S. (2025, September). RSEs 2035: surviving or thriving in the age of AI. In 2025 IEEE International Conference on eScience (eScience) (pp. 381-382). IEEE. He, H., Miller, C., Agarwal, S., Kästner, C., & Vasilescu, B. (2025). Speed at the cost of quality: How Cursor AI increases short-term velocity and long-term complexity in open-source projects. arXiv preprint arXiv:2511.04427. Jarrahi, M. H. (2018). Artificial intelligence and the future of work: Human-AI symbiosis in organizational decision making. Business Horizons, 61(4), 577-586. O'Brien, G. (2025). Threats to scientific software from over-reliance on AI code assistants. Nature Computational Science, 5, 701–703. O'Brien, G. (2025). How scientists use large language models to program. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (pp. 1-16). ACM. O’Brien, G., & Eisty, N. U. (2026). Scientific software in the age of vibe coding. Computing in Science & Engineering, 28(1), 112-116. Ribes, D., Jackson, S., Geiger, S., Burton, M., & Finholt, T. (2013). Artifacts that organize: Delegation in the distributed organization. Information and Organization, 23(1), 1-14. Song, F., Agarwal, A., & Wen, W. (2024). The impact of generative AI on collaborative open-source software development: Evidence from GitHub Copilot. arXiv preprint arXiv:2410.02091. Xu, F., Medappa, P. K., Tunc, M. M., Vroegindeweij, M., & Fransoo, J. C. (2025). AI-Assisted programming decreases the productivity of experienced developers by increasing the technical debt and maintenance burden. arXiv preprint arXiv:2510.10165.