Agentic artificial intelligence in microbiology: opportunities, challenges, and future directions
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Agentic artificial intelligence (AI) systems extend beyond task-specific machine learning by iteratively integrating perception, reasoning, planning and action within adaptive feedback loops. In microbiology, where discovery increasingly depends on multi-omics integration, high-throughput experimentation and complex ecological modeling, such systems could transform how hypotheses are generated, prioritized and experimentally tested. However, agentic AI introduces not only technical innovation but epistemic and institutional challenges. Iterative autonomy may amplify error propagation, reshape authority in knowledge production, and increase systemic fragility if not bounded by robust oversight and provenance-aware design. In this Review, we distinguish automation from agency and outline operational architectures for agentic AI in microbiology, including stratified autonomy levels, policy-constrained tool invocation, digital twin testing environments and persistent scientific memory grounded in provenance standards such as W3C PROV. Using antimicrobial resistance as a case study, we illustrate how agentic loops could enable adaptive exploration while minimizing experimental redundancy. We then analyze epistemic risks, including bias amplification, automation bias and reproducibility drift, and propose governance principles that embed resilience into system design. We argue that the responsible integration of agentic AI in microbiology depends not solely on performance gains but on explicit autonomy boundaries, uncertainty calibration, auditability and institutional alignment. Rather than replacing microbiologists, agentic AI should be understood as a new class of scientific instrumentation whose legitimacy rests on transparency, accountability and epistemic robustness.
Fields of Study & Subjects
Artificial Intelligence Applications
Scientific Computing and Data Management
Ethics and Social Impacts of AI
Publication details
- Journal / source
- HAL
- Publication date
- 17 February 2026
- Language
- EN
- Document type
- Preprint