AI agents move into biomedical research workflows

AI Agents are emerging as a new layer of scientific software, using LLM-based systems to coordinate tools, delegate tasks and manage complex workflows beyond a chatbot interface. Researchers at CU Anschutz describe multi-agent systems as networks of specialized components, often led by an orchestrator that routes work to agents focused on literature review, hypothesis generation, ranking, coding or data analysis.

Public interest in AI Agents has risen more than 550% from the previous year, according to Google Trends cited by the researchers, but the concept is not new. Yanjun Gao notes that earlier agent systems used techniques such as slot filling to handle tasks like travel booking, while newer implementations rely on LLMs and tool calling to search databases, execute Python code, interpret files or query scientific resources such as PubMed and OpenAlex.

Biomedical teams are testing these systems as research assistants rather than replacements for scientists. The Pividori Lab is using agents in Type 1 diabetes work to explore gene modules, molecular patterns and possible hypotheses, while Gao’s LARK Lab sees potential for supporting cross-disciplinary research. Researchers also warn that healthcare use must follow institutional rules, HIPAA requirements and data regulations, and that a simpler chatbot may be more appropriate than a complex agentic system for many tasks.