What is guardrails?
Rules and checks that keep an AI agent within safe, approved behaviour.
Guardrails, explained
Guardrails can restrict topics, block certain statements (such as medical or financial advice), enforce politeness and trigger escalation.
On phone calls, guardrails also cover tone and timing: how to respond to an angry caller, when to stop calling someone, and which topics always go to a person. Clear guardrails let businesses use AI on calls without losing control of what is said.
See how these ideas come together in Chitti, Manolle's AI voice agent.
Related terms
- Hallucination (in AI)When an AI states something that sounds plausible but isn't true.
- EscalationPassing a conversation to a human or a higher level of support when the agent shouldn't or can't handle it.
- System promptThe instructions that set a voice agent's role, tone, rules and goals.
See an AI voice agent at work on your calls.
See the complete system live: marketing, lead capture, AI voice calling, sales conversion, and analytics.