AI SDR Conversation Moderation is a built-in safety feature of Piper Conversations. This feature is designed to assist in moderating conversations and mitigating exposure to inappropriate content. It is designed to automatically identify and end conversations that contain inappropriate or harmful content, helping maintain a professional representation of your brand. Enforcement is typically immediate upon detection of a policy violation.
This allows your AI SDR agents to stay on topic and on message, focusing on delivering pipeline and engaging serious buyers.
How it works
Our system performs real-time analysis to assess a visitor's message content as it's sent. A dedicated AI moderation agent scores the message across several policy categories. If a message's score exceeds a predefined safety limit in any category, the conversation is ended.
High-risk policy categories monitored
The AI moderation agent targets content within several categories. Monitored categories include, but are not limited to:
-
Sexual content.
-
Harassment (direct insults or threats).
-
Hate speech (targeting protected attributes).
-
Violence.
-
Self-harm.
-
Illegal activities.
-
Personal financial distress
-
Legal advice
The AI moderation agent also includes heuristics to detect and respond to common attempts at jailbreaking or prompt hacking and will end the conversation if these behaviors are detected.
What happens when a conversation is ended for high-risk conversation topics
If a visitor's message violates our content policies, the enforcement is immediate and automatic.
-
The visitor will no longer be able to send messages in that conversation.
-
They will see a system message stating: "Conversation ended due to inappropriate content."
-
To reduce the visibility of harmful content, conversation history is restricted within the Qualified app for non-admin users.
Reporting and analytics
For your records, all moderation actions are logged. As an Admin, you can easily find this data using the “AI SDR ended conversation?” filter. This filter is available in both your standard reports and in custom dashboards, allowing you to track the frequency and context of moderated conversations.
Off-topic content detection
Not all off-topic messages are a concern. If a visitor asks about topics not relevant to your business or sends gibberish, the agent can simply clarify it's not the right place and move on - no harm done. If a visitor is confused about a wrong topic or service, asks for harmless information or makes small talk, or speaks in gibberish, the agent will redirect the conversation back to your products and services, but will not flag or end the conversation for harmful activity.
.svg)
