reply-classification

What problem does it solve?

Manually sorting through inbound email replies is slow, inconsistent, and risky for high-volume workflows like sales outreach, customer support, and transactional email. Failing to quickly identify hot leads, honor unsubscribe requests, or handle bounces can lead to lost revenue, CAN-SPAM compliance violations, and damaged sender reputation.

Core Features & Use Cases

• Proven 9-category intent taxonomy: Maps common reply types (interested, not interested, out-of-office, bounce, unsubscribe, question, support, billing, unclassified) directly to actionable routing rules.
• Near-perfect auto-reply and bounce detection: Uses RFC-standard header checks to identify automated responses and delivery status notifications without analyzing message content.
• Weighted keyword intent scoring: Classifies non-automated replies with subject-weighted matching, density adjustment, and confidence thresholds to reduce false positives.
• Built-in safety and compliance guardrails: Parallel safety classification blocks phishing and malicious content, while rules automatically honor unsubscribe requests and escalate high-risk intents like legal or security queries.
• Use cases: Prioritize hot sales leads, triage support inboxes, manage transactional email opt-outs, and reduce manual triage time for high-volume inbound email flows.

Quick Start

Use the reply-classification skill to categorize the most recent inbound reply to your sales outreach email, then route it to the appropriate team or queue based on its classified intent and confidence level.