Evolving the RSE role in the age of generative and agentic AI

Generative AI and agentic tools (GenAI) are dramatically changing research software practice. Researchers can now generate their research code more easily with less programming and software engineering experience. Anecdotal evidence suggests that researchers’ collaborations with RSEs now focus on more complex or higher-level software engineering tasks such as code review.

These changes bring uncertainty: Will the RSE role be reduced to reviewing and refactoring AI-generated codebases, or combining prompt engineering and configuring agentic workflows? How can we assert the value of RSEs to our institutions, funders, and collaborators in the future? Will agentic systems replace RSEs and RSE groups?

We present a vision for the future of the RSE role and its value proposition, developed at a recent ReSA workshop. We believe that while the RSE role will change with GenAI, its value remains because of RSEs’ unique insight into the interplay between research methods and software engineering. RSEs combine deep understanding of research disciplines with professional software engineering expertise, which is critical to the collaborative effort of research in assuring that sustainable research software supports the highest quality reproducible research. To realise this value, the RSE role will increasingly integrate expertise in relevant GenAI methods and tools.

We summarise our discussion and propose a roadmap for future collaboration within the research software community to develop measures for future-proofing research software engineering and the RSE role. We also highlight early outcomes from a Birds-of-a-Feather session on this topic, held at the 1st International Research Software Conference.