Skip to main content

​
Overview

Agent Injection enables users to construct custom AI agents within workflows that function as team members in conversations, powered by Conduit’s AI infrastructure.
Key capability: Works with both text and VoiceAI channels.

​
Use Cases

The node supports three primary scenarios:
  • Guide users through multi-step processes
  • Collect missing information from contacts
  • Execute repeated follow-ups until objectives are achieved
Users maintain full control over agent tone, logic, and follow-through behaviors.

​
Implementation Steps

1

Initial Setup

Navigate to your workspace, create a trigger, then select Add Action > Agent Injection (or Agent Injection Phone for voice).
2

Agent Identity

Assign a recognizable name (example: “Carl from Conduit”) to establish agent personality.
3

Agent Description

Define behavioral guidelines with a concise prompt such as: “You are Alex, a friendly customer support representative.”
4

Role Selection

Choose one of two modes:
  • Action: Persuade users toward specific outcomes (booking confirmations, form uploads)
  • Information: Extract particular data points (delivery ETAs)
5

Task Configuration

  • Write detailed instructions covering tone, constraints, and fallback protocols
  • Set follow-up frequency and intervals for retry attempts
6

First Message

Configure initial outreach by selecting:
  • Communication channel (Email, SMS, WhatsApp)
  • Sender identity
  • Subject line (email only)
  • Message body with optional dynamic variables (e.g., {{guest.first_name}})
Once deployed, the AI manages subsequent replies until task completion or follow-up limits are reached.
Important Note: Agent Injection works across multiple inboxes with format-adaptive behavior adjustments.