Reply Classification

Automatically classify inbound replies into actionable categories. Not all replies are equal — an out-of-office shouldn't stop your sequence, but a "not interested" should.

Classification Categories

interested       — Positive signal, route to human follow-up
not_interested   — Explicit decline, stop sequence and suppress contact
out_of_office    — Automatic vacation reply, pause and resume
wrong_person     — Not the right contact, stop sequence
unsubscribe      — Wants to be removed, suppress contact permanently
auto_reply       — Delivery receipt or auto-ack, ignore and continue

How It Works

When an inbound message is received, OutreachAgent classifies it using an LLM (OpenAI gpt-4o-mini when configured) with a heuristic fallback. The classification is:

• Stored on the message as classification
• Included in the message.received webhook event payload
• Available for classification-aware exit criteria

Classification-Aware Exit Criteria

Extend exit criteria with replyClassifications to control which reply types trigger an exit:

{
  "exitCriteria": [
    {
      "trigger": "reply",
      "replyClassifications": ["interested", "not_interested"]
    },
    { "trigger": "bounce" },
    { "trigger": "unsubscribe" }
  ]
}

With this configuration, out_of_office and auto_reply responses will not stop the sequence.