AI may change how research software is built — but not why it matters
In March 2026, UCT eResearch participated in an international workshop in Edinburgh that brought together leaders from the global research software and AI communities to explore how generative AI is changing the way research software is developed. Research Software Engineering in the Age of Generative AI: Building a Community Vision was co-led by the Research Software Alliance who was funded to undertake this work as part of a Schmidt Sciences grant. The discussions ranged from software verification and training to governance, sustainability, and the future role of Research Software Engineers (RSEs).
While much of the global conversation about AI focuses on technological capability, one message was specifically highlighted: the challenge is no longer simply producing software but ensuring that it can be trusted.
AI-assisted development tools can already generate code, documentation, tests, and workflows at unprecedented speed. Yet research depends not on how quickly software is produced, but on whether the resulting outputs are reliable, reproducible, and scientifically meaningful. As software creation becomes easier, verification, validation, and transparency become increasingly important.
For African universities, this raises important questions.
- Will AI reduce barriers to computational research, allowing more researchers to develop software and participate in digital scholarship? Almost certainly.
- Will access to AI tools alone solve challenges around research software sustainability, training, support, infrastructure, and research quality? Almost certainly not.
- Will access to AI tools and infrastructure be equally accessible? Definitely not.
The Edinburgh discussions reinforced the idea that investment in people remains critical. Research software expertise, training, governance, and communities of practice will continue to be essential if institutions are to realise the benefits of AI while managing its risks.
This is particularly relevant in Africa, where universities are simultaneously expanding their digital research capabilities and navigating uneven access to infrastructure, funding, and specialised technical expertise. AI has the potential to accelerate research, but it may also amplify existing inequalities if access to tools, training, and support remains concentrated in well-resourced institutions.
For UCT, the conversation is therefore not simply about adopting AI tools. It is about building the capabilities, practices, and support structures needed to use those tools responsibly in research.
The future of research software will not be determined by AI alone. It will be shaped by the choices that institutions, researchers, and research support professionals make about how these technologies are integrated into research practice. The opportunity for African universities is not only to adopt emerging approaches, but also to help shape them.
Outputs from the workshop are available online:
- Workshop report - Research Software Engineering in the Age of Generative AI: Building a Community Vision (https://doi.org/10.5281/zenodo.20320884)
- Draft vision statement/discussion document - Research Software in an Age of AI-Assisted Development: Reflections from Edinburgh: (https://doi.org/10.5281/zenodo.20321134)
- Blog post - Research Software Engineers in the Age of GenAI: Same Value, Changing Practice (https://www.researchsoft.org/blog/2026-05-28/)