An Agent on a Leash, or why my AI agent doesn't make business decisions

This post kicks off an ongoing experiment: building an LLM-powered support agent you can actually...