AI Activity Log

A complete record of everything your AI agent has said. Stay informed, catch issues early, and improve your agent over time.

Viewing AI Activity is available on all plans. AI Auto-Reply requires Growth or Scale.

Overview

The AI Activity Log gives you a clear view of every automated reply your SMS agent has sent. Whether your agent handled 10 conversations or 10,000 overnight, you can review exactly what it said, to whom, and when — without digging through individual conversation threads one by one.

Use the AI Activity Log to stay informed, verify that your agent is behaving as intended, and identify opportunities to improve your system prompt.

How It Works

Accessing the log

In your Conversations inbox, click the AI Activity tab. This filters your conversation list to show only the conversations where your AI agent sent at least one reply.

Select any conversation in the list to open the full thread. AI-generated messages are clearly labeled within the thread so you can distinguish them from messages sent by a human team member.

What the log shows

For each AI-assisted conversation, you can see:

• The contact's name and phone number — who the agent was talking to.
• The full message thread — every message from the contact and every reply the agent sent, in order.
• Timestamps — when each message was sent and received.
• Escalation status — whether the agent flagged the conversation as Needs Human Attention and at what point in the conversation that happened.

Spotting issues

If your agent gives an unexpected, unhelpful, or off-brand response, the AI Activity Log is where you will find it. Look for conversations where the contact seemed confused, stopped responding, or where the agent repeated itself or failed to answer a direct question.

These patterns are signals that your system prompt needs refinement. See Creating an SMS Agent for guidance on updating your instructions.

Tips & Best Practices

• Review the log after every major broadcast. When a large broadcast goes out and contacts start replying, the AI agent handles the initial wave. Check the log within 24 hours to catch any systematic issues before they affect more contacts.
• Pay attention to escalations. When and why your agent escalates tells you a lot about the gaps in your instructions. If it escalates too often, your system prompt may be too restrictive. If it never escalates, it may be trying to handle things it should not.
• Look for patterns, not just outliers. One unusual response could be a fluke. Several contacts getting the same confusing reply is a signal that your prompt needs to be updated.
• Use good responses as benchmarks. When you see a conversation where the agent handled things particularly well — qualified a lead, answered objections, booked a callback — note what the contact said and how the agent responded. That context can help you refine your prompt for similar scenarios.
• Share notable conversations with your team. The AI Activity Log is useful for training purposes. A team member reviewing real AI conversations can quickly understand what the agent does well and where humans need to step in.