Reply Intelligence: Automatic Classification of Replies

When a lead responds to one of your campaign emails, Leadterra captures the full reply thread and runs it through an automatic intent classifier. Rather than dumping raw email text into your workflow, the platform hands you a structured classification — interested, not interested, out of office, and more — so your agents and downstream systems can take the right action immediately without building their own NLP layer.

Reply classification

Every inbound reply is analyzed and assigned one of the following classifications:

Classification	Meaning
`interested`	The lead expressed positive intent or asked to learn more.
`not_interested`	The lead declined or asked to stop outreach.
`out_of_office`	An automated away message with no actionable intent signal.
`bounce`	A delivery failure notification returned to the sender.
`unsubscribe`	The lead explicitly requested removal from your list.

Classifications let your automation branch immediately — for example, routing interested replies to a human SDR queue, suppressing unsubscribe leads from future campaigns, and ignoring out_of_office replies without clogging your pipeline.

Pulling replies

Use GET /v1/replies to list classified replies across all campaigns, or filter to a specific campaign with the optional campaignId query parameter:

curl "https://app.leadterra.co/v1/replies?campaignId=camp_01hx4mrqpbcn7wh2gfk9yz6e3s" \
  -H "Authorization: Bearer sk_live_YOUR_KEY"

Sample response

{
  "data": [
    {
      "id": "rpl_01hx5ntrqbdp8xj3ghm2az7c4k",
      "email": "priya.sharma@voltmetrics.io",
      "campaignId": "camp_01hx4mrqpbcn7wh2gfk9yz6e3s",
      "classification": "interested",
      "receivedAt": "2025-01-15T11:04:32Z",
      "threadSnippet": "Hi James, yes this looks relevant — can we find 20 minutes this Thursday?"
    },
    {
      "id": "rpl_01hx5ntrqdep9yk4him3ba8d5l",
      "email": "dan.foster@loopcraft.com",
      "campaignId": "camp_01hx4mrqpbcn7wh2gfk9yz6e3s",
      "classification": "not_interested",
      "receivedAt": "2025-01-15T13:47:09Z",
      "threadSnippet": "Thanks but we're all set for now. Please remove me from your list."
    }
  ],
  "meta": {
    "total": 2,
    "page": 1,
    "pageSize": 50
  }
}

The threadSnippet field gives you enough context to act on the reply without fetching the full thread. Use the email field to look up the lead record and classification to route the reply to the right handler in your workflow.

Using replies in agent workflows

Replies are the signal that closes the outbound loop. Pull GET /v1/replies on a schedule or trigger downstream logic from the reply.received webhook (see below), then branch your agent’s behavior on classification:

Pass interested replies to a Deepline agent or Claude Code workflow that drafts a personalized follow-up or books a meeting.
Feed not_interested and unsubscribe classifications into a suppression list so those leads are excluded from future campaigns.
Ignore out_of_office replies or log them and retry the lead after a configurable delay.

Because the reply object includes email, campaignId, and threadSnippet, your agent has all the context it needs to take action without additional API calls in the common case.

Webhook alternative

Instead of polling GET /v1/replies, you can subscribe to the reply.received webhook event. Leadterra will POST the reply payload to your endpoint the moment a new reply is classified, giving you a real-time push model with no polling overhead. The webhook payload mirrors the reply object returned by GET /v1/replies, so your handler code can be the same regardless of whether you use polling or push.

Use the reply.received webhook rather than polling GET /v1/replies whenever your workflow needs to act within seconds of a reply arriving. Polling introduces latency proportional to your polling interval and adds unnecessary API calls during quiet periods. Reserve polling for batch reconciliation or backfill scenarios.

Get Started

Core Concepts

Guides

Reply Intelligence: Automatic Classification of Replies

Reply classification

Pulling replies

Using replies in agent workflows

Webhook alternative

​Reply classification

​Pulling replies

​Using replies in agent workflows

​Webhook alternative

Reply classification

Pulling replies

Using replies in agent workflows

Webhook alternative