Inbox & AI Replies
The unified inbox
Every reply - LinkedIn message, email, social - lands in one inbox, grouped by conversation thread regardless of channel. When a contact replies, the AI drafts a response in the background; you see the draft alongside the original message and can approve, edit, or reject it before anything is sent.
Training the AI’s voice
Under AI Assistants, a Communication Profile controls how the AI writes: tone (friendly, formal, direct, casual, or technical), language, example messages to imitate, your product description and value proposition, and explicit rules (“never mention pricing,” “always ask for a 15-minute call”).
Don’t want to write one from scratch? The Voice Library has 4 starting templates - Direct & Technical, Warm & Consultative, Formal Executive, and Founder-to-Founder - you can pick and then customize.
Escalation - when the AI should NOT reply
Set escalation triggers: specific situations (a pricing question, a complaint, a request to speak to a human) where the AI stops and flags the conversation for you instead of drafting a reply. This is the main safety valve - use it for anything you don’t trust an AI to handle correctly on its own.
Require approval before sending is a per-campaign setting. With it on (the default), every AI draft sits in your inbox until you approve it - nothing goes out automatically. Turn it off only once you trust the assistant’s judgment for that specific campaign’s audience.
Testing before you go live
The Sandbox lets you send test messages to your AI assistant and see exactly how it would respond, using your real Communication Profile - without touching any real contact or campaign. Use it to catch a bad tone or a wrong fact before it reaches an actual prospect.
AI usage
AI replies (plus rewrite/regenerate/ICP generation/chat) count against your plan’s monthly AI actions allowance - see Billing & Plans for the exact numbers per tier. Usage is visible on the Billing page as you go, not just at month-end.