Why AI agents might not be right for you

Nondeterministic LLM-based agents can give different outputs for the same input, so they suit tasks with changing inputs rather than single right answers.

The article says agents are a poor choice where deterministic software suffices, where every action must be exactly correct, where misbehavior could cause serious harm, or where their cost exceeds the expected benefit.

It advises setting an error budget, a site reliability engineering concept, to decide whether to deploy an agent.