Originally published on robatdasorvi.com
Last month I watched a customer support agent refuse to answer a question it had already answered three messages earlier. A second agent in the same pipeline had flagged the topic as sensitive. The first read the flag, deferred to it, and apologised. Neither agent was wrong by its own rules. Together they produced an outcome nobody intended.
I keep thinking about that one. Nobody's system failed. That was the problem.
Why agent rules conflict in multi-agent pipelines
Agent rules conflict because each agent carries constraints from a different source, and those constraints were written without knowledge of each other.
