Intelligence

Inbound Classification for AI Agents

Every reply is classified by intent with a confidence score and suggested action. Your agent knows exactly what to do next — no rules engine required.

How it works

1

Receive an inbound email

Replies to emails your agent sent are automatically captured and linked to the original thread. You can also manually record inbound emails via POST /v1/agent/record-inbound.

2

Classify the intent

POST to /v1/agent/classify-intent with the subject and body text. The system returns an intent label, confidence score, and suggested action.

3

Route based on intent

Suggested actions tell your agent what to do next: notify the owner, require human approval, auto-archive, or escalate. Your agent decides how to act on the recommendation.

4

Batch processing

POST to /v1/agent/batch/classify-intent to classify multiple messages in a single request. Each message gets its own intent, confidence, and action.

5

SLA tracking and threading

Classified replies are linked to their conversation thread. Thread statuses transition automatically — open, waiting, resolved, escalated — enabling SLA tracking.

Classify an email

POST /v1/agent/classify-intent

POST /v1/agent/classify-intent
{
  "tenantId": "your-tenant-id",
  "subject": "Re: your proposal",
  "bodyText": "Thanks, I'd love to schedule a demo next week."
}
{
  "intent": "interested",
  "confidence": 0.92,
  "suggestedAction": "notify_owner"
}

Intent categories

Interested

Positive engagement — the contact wants to proceed, book a demo, or learn more.

Objection

Concern or pushback on the offering. May need a tailored follow-up.

Support

Needs help with the product or service. Route to support workflows.

Billing

Payment, invoice, or subscription related. Route to billing handlers.

Out of Office

Auto-reply or OOO message. Safe to archive without action.

Not Now

Timing isn't right but may revisit. Schedule a follow-up for later.

Batch classification

Classify multiple messages in a single request. Each message gets its own intent, confidence score, and suggested action.

POST /v1/agent/batch/classify-intent
{
  "tenantId": "your-tenant-id",
  "messages": [
    { "subject": "Re: demo", "bodyText": "Yes, interested" },
    { "subject": "OOO", "bodyText": "I am out of office until Monday" },
    { "subject": "Re: invoice", "bodyText": "Can you resend the invoice?" }
  ]
}

Understand every reply

Automatic intent classification with confidence scoring and suggested actions. Let your agent triage replies at scale.