#multi-agent

What Breaks When You Split a Job Across AI Agents

Five LLM agents worked a bank fraud case in sequence, each passing its finished work to the next. In the natural setup, one agent beat every chain. What helped chains most was giving each agent a list of fields to fill in. Free text with the same field list as the JSON records completed 49 of 60 cases, against 17 without it, and with that list a chain matched one strong agent. Free-text chains that failed had dropped details and added conditions nobody could meet. A code without its definition stalled payments, and wrong work no later agent could check went straight through.