Skip to content

Explanations

The rest of this site says how. These pages say why: what agents are changing about the work, what the evidence shows, and what it might mean for curation. They say where the evidence is thin.

Page Question it answers
How agents changed software development What happened to developers' roles and processes in 2026, the jagged frontier, and which components to craft by hand versus generate
What other fields suggest about curation What radiology and evidence synthesis show about second readers and the costs of human review, and what that might mean for biocuration

A guide for discussion

These are questions GO curators and developers have raised about using AI, with the section that addresses each.

Question Where to read
Is adoption too slow or too fast? Expectations versus reality Felt speed and measured speed differ; The jagged frontier
The developers' perspective How roles change
Project management, information overload, multitasking Supervision is tiring; What we do in practice
Are we more efficient? Can we do more with less? Could it be much more efficient?
Is the code and data we produce of higher quality? Craft the core, generate the periphery; Second eyes in curation
What does a human in the loop cost? Do experts have to be hypervigilant? The costs of keeping a human in the loop
How do we spot the output most likely to be wrong? Where to look
Can AI review AI? What benchmarks could we use? AI reviewing AI; How to tell whether AI review works
How should skills development be coordinated? Write your own skills; Writing and sharing skills on the GO AI Hub; Break work into skills
How do curators share use cases and workflows? Practical starting points; GO AI Hub

Talks

These pages build on talks given to the curation community since late 2025. Slides are on Zenodo.

  • Staying in the Loop: A Biocurator's Guide to Agentic AI Developments. Biocuration 2026 virtual conference, February 2026. Slides.
  • Challenges with deploying agentic AI for biomedical data integration and curation. AIBIO-UK, EMBL-EBI, February 2026. Slides.
  • Using agentic AI to assign gene function across multiple genomes. ASBMB 2026, March 2026. Slides.
  • Agentic AI and Biodata Resources: The need for action. Global Core Biodata Resource Forum, October 2025. Slides.
  • How Agentic AI can help us in GO and Using Agentic AI to Review GO Standard Annotations. GO Consortium meeting, October 2025. Part 1, part 2.
  • GO curators agent training, April 2026, the workshop described in the GO AI Hub case study. Materials.