The best AI agents are simpler than you think
Sierra builds customer-engagement agents using many parallel models per turn and isolated PCI infrastructure for payments.
“We have isolated infrastructure where payment info doesn't go to an external large language model cuz none of the LLM providers are PCI certified in that way.”
Sierra's head of product details how their customer-engagement agents invoke 10-15 models per conversation turn to parallelize thinking, listening, and talking, and run a separate PCI-isolated payment layer outside any LLM. It matters because it shows production agent architecture diverging sharply from standard single-LLM-call harnesses, including emerging agentic commerce where agents earn sales commissions.