What Is the AI Reply Agent?
The AI Reply Agent is the first AI Agent we launched. It is like an automated teammate for your inbox. It reads incoming emails from leads, replies on your behalf, handles objections, sends follow-ups, shares calendar links, and even updates the CRM status, all without needing constant oversight.
Why it matters: If you’re sending thousands of emails per day, you’re likely getting hundreds of replies. Some leads want more info. Some are ready to book. Others raise concerns. Manually managing all of that is time-consuming and inconsistent.
Speed to lead is crucial because the faster you respond to a prospect, the more likely you are to convert them.
The AI Reply Agent ensures replies go out instantly (under 5 minutes), 24/7, keeping leads warm, answering questions, and booking calls while intent is still high, without relying on your team to be constantly available.
How to Set It Up
From the left sidebar, navigate to the AI Agents section and click “Add New.” Name your agent, then choose AI Reply Agent.
Choose how hands-on you want it to be. You can use Human-In-The-Loop (HITL) mode, where you approve replies before they’re sent. Or go fully automated with Autopilot, where the agent replies to everything instantly. The second option is ideal for experienced teams who’ve already fine-tuned their workflows.
Configure your agent’s behavior. Decide whether it should handle follow-ups, respond to out-of-office replies, and engage with objections or negative responses.
Then, assign the campaigns or inboxes it should monitor. You can set it as a default agent for all campaigns, or assign specific campaigns and accounts.
Train the agent on your tone of voice and key product details. This step is crucial; the more context you provide, the more useful and accurate the replies will be.
Click Optimize to help refine your guidance by checking for common issues and suggesting improvements.
Here are some examples of what guidance can look like:Your goal is to qualify and convert interested replies into booked calls.
When someone shows interest or asks for more info:
- Acknowledge their interest clearly and politely.
- Ask a simple qualifying question if needed (e.g., "Are you targeting outbound for your own business or running it for clients?")
- Suggest a call to go deeper and explain what they'll get from it.
Booking Prompt Example:
“Sounds like this could be a good fit. Let’s do a quick call to see how we can help. Here’s the link to book a time: [Calendly Link]”
If they ask specific questions (e.g., pricing, deliverability, how it works):
- Answer clearly and briefly.
- Still suggest a call for full context.
Always keep replies:
- Direct and friendly
- Focused on value and next step (call)
- Avoid fluff, no salesy phrases
Example Style:
“Yes, we include lead sourcing and full campaign setup. Most clients see results within 3–4 weeks. Let’s hop on a quick call so I can walk you through the setup: [Calendly Link]”
Don’t:
❌ Use exaggerated claims (“insane results,” “game-changing”)
❌ Write like a robot (“Thank you for your message. I am happy to assist you.”)
❌ Be vague (“It depends,” “We’ll see what works”)
❌ Sound promotional (“This is the best solution out there!”)
Autopilot vs. HITL
If you’re new to automated replies, start with Human-in-the-Loop. It gives you control and helps you fine-tune tone and responses before fully handing it off.
If you’re managing high volume or need full automation, Autopilot is your go-to. Once trained, it replies instantly without needing your input, perfect for scaling!
Tracking replies (AI Agent Details)
You can track all AI Agent replies and interactions in the Activity Feed, which gives you a real-time view of every conversation. It shows when an email was sent, when a reply needs review, or when the AI requested more info.
Each item is time-stamped and color-coded by action, making it easy to spot what’s been handled and what still needs attention. From here, you can also take action, like reviewing drafts, marking conversations as done, or digging into specific lead replies.
1. Involvement Rate
The Involvement Rate shows how often your AI agent participates in email conversations. It measures the percentage of email threads where the AI agent was actively involved in helping craft responses.
The AI is considered "involved" whenever it generates a draft reply, regardless of whether you send it as-is, modify it, or even decide not to send it at all.
As long as the AI generated a suggestion for you to review, it counts as involvement.
This includes both cases where you send the AI's draft unchanged and cases where you edit the draft before sending.
The only time the AI is NOT considered involved is when you completely skip its suggestion and write an entirely manual response from scratch.
Why it matters: A high involvement rate indicates that your AI agent is successfully identifying and responding to conversations that need attention, saving you time by providing draft responses.
2. Resolution Rate
The Resolution Rate measures how often the AI agent can handle email responses completely hands-free, meaning you can send the AI's draft without any edits or intervention.
This metric only counts emails that were actually sent (not pending drafts).
A "hands-free resolution" occurs when you send the AI's draft exactly as generated, without making any changes.
If you edit the draft in any way before sending, it doesn't count toward the resolution rate.
If you skip the AI's suggestion entirely and write your own response, this also doesn't count as a hands-free resolution.
Why it matters: A high resolution rate means the AI is accurately understanding and responding to emails in a way that matches your communication style and needs, requiring less manual intervention from you. This directly translates to more time saved.
Both metrics work together to show you how effectively your AI agent is helping manage your inbox - the Involvement Rate shows how often it's helping, while the Resolution Rate shows how accurately it's helping.
FAQ
Q: What does the AI Agent's Calendly integration do exactly?
A: The Calendly integration allows the AI Agent to send your booking link directly in the email reply when a lead expresses interest. It can also mention available time slots, if helpful, but the lead needs to initiate the booking by clicking the link. Fully automated booking (where the agent books a slot on the lead’s behalf) is not available yet, but it's on the roadmap.
Q: How does routing work when handover or escalation is triggered?
A: When the AI Agent decides to route a conversation to your team, it notifies you in three ways:
Slack notification so you’re alerted in real time
The lead appears in the Activity section, where you can review the message, respond, or mark it as done
The conversation is also visible in the Unibox, alongside all other replies, for full context and follow-up
This ensures no escalated lead gets missed.
Q: Will the Autopilot AI Agent always send replies automatically without needing my approval?
A: Yes, Autopilot is designed to send replies automatically, but only when the agent is confident in its response. If it’s unsure (for example, due to missing product info or unclear context), it pauses and asks for your review instead. You can check the reason by opening the agent, going to the Activity section, finding the specific lead, and reviewing the message details. It will show why the reply wasn’t sent automatically.