Are artificial intelligence (AI) agents ready for medicine and biomedical research? A narrative review.

English summary

This narrative review describes how large language models are being extended from single-turn text generation into artificial intelligence (AI) agents that pursue goals through planning, retrieval, tool use, code execution, memory, feedback, and role coordination. In medicine and biomedical research, early applications include clinical calculations, risk prediction, oncology decision support, omics analysis, hypothesis development, laboratory automation, and research writing.

Key points

• The most defensible clinical applications are those using validated calculators, curated clinical tools, or guideline-grounded modules under human oversight.
• Biomedical discovery systems support broader workflows, but many claims still depend on preprints, narrow benchmarks, simulated settings, or domain-specific demonstrations.
• The immediate task is not to determine whether AI agents will replace experts, but to identify which tasks can be delegated, what evidence is required, and which forms of human judgment must be preserved.

Audience

This summary is intended for clinicians, biomedical researchers, medical AI and data-science researchers, and advanced students.

Scope and limitations

The supplied text presents the main points of a narrative review but does not provide detailed search methods, eligibility criteria, quantitative evaluations, or full validation evidence for individual examples. Accordingly, the strength of evidence is uneven across applications, and claims concerning biomedical discovery may rely on limited forms of evidence.

Research context

The article connects with medical AI and data science through its focus on agent-based systems in clinical and biomedical workflows, including evaluation, reproducibility, human oversight, and accountability.