Talk to Jade
// knowledge base

Find answers by category.

Browse categories on the left and open articles in the main panel.

← Back to Security, Privacy & Compliance

AI Guardrails: Preventing Hallucinations

Keeping the AI on Script

One of the biggest concerns enterprise clients have about AI is the risk of "hallucinations"—when an AI makes up facts or gives incorrect advice. We prevent this through a highly controlled architecture.

Strict System Prompts

Your agent does not use the open internet to answer questions. Instead, it relies strictly on the isolated knowledge base we build during your onboarding process. We apply rigorous "guardrails" in the agent's system instructions. If a caller asks a question that is not covered in your approved documents, the AI is explicitly instructed to politely state that it does not have the information and will immediately offer to book a call-back with a human team member.

This ensures your brand reputation is protected, and callers only ever receive factual, verified information.