Why agent settings matter
Agent settings shape the caller’s first impression and every interaction that follows. The greeting is the first thing a caller hears — it goes directly to text-to-speech without LLM processing, so it needs to be exactly right. The personality and role inform how the agent communicates, while rules enforce consistency and compliance across every conversation.When to configure each setting
| Setting | When to use | Impact |
|---|---|---|
| Agent | During initial setup, or when rebranding or changing your agent’s purpose | Controls the greeting line (first words spoken), personality (tone), and role (job function). Directly shapes caller perception. |
| Model | When standing up a new agent, or when optimizing for cost, latency, or reasoning quality | Determines the LLM backbone. PolyAI’s Raven models are recommended for most deployments — they are purpose-built for conversational AI and offer superior grounding and natural speech. Third-party models are also available. |
| Rules | When you need consistent terminology, regulatory compliance, or defined handling of edge cases | Sets global behavioral constraints — tone, language style, compliance guardrails, and pronunciation overrides that apply to every conversation. |
Configuration sections
Agent
Set the greeting line, personality, and role that define how your agent introduces itself and communicates.
Model
Choose your LLM — PolyAI’s Raven (recommended), OpenAI, or Amazon Bedrock — or connect your own endpoint.
Rules
Define global rules for language style, compliance, edge-case handling, and TTS pronunciation.