reply-engine

Automating LinkedIn message classification requires a system that categorizes inbound threads and generates contextual responses. This Skill handles classification, automated replies, and thread management across LinkedIn and email, including proposing meetings or marking dead conversations.

Human review triggers flag messages needing manual attention due to high confidence scores, sensitive content, or high-value leads. This ensures critical conversations bypass automated responses and reach human operators for personalized handling.

Managing conversation threads requires a database with a specific schema supporting message classification and thread context. You need PostgreSQL configured via SQLAlchemy and psycopg2 to store thread states, lead scores, and response history.

A/B testing for automated responses evaluates different reply variations against classified message categories. The system handles various scenarios including meeting proposals and referral handling, comparing response effectiveness across classified thread contexts.

Slack integration for lead scoring notifications uses the Slack SDK to alert teams about classified messages. This enables real-time monitoring of inbound LinkedIn and email threads, especially for high-value leads flagged for human review.

Handling referrals and meeting proposals in conversation management requires classifying message intent and generating contextual responses. This Skill manages thread states by proposing meetings, handling redirects, and marking unresponsive conversations as dead.