How Contactly's AI inbox works

Every submission that lands in Contactly’s shared inbox goes through the same pipeline before anyone opens it.

1. Spam scoring

Each submission is checked against honeypot fields, time-trap signals, rate limits, and a spam-scoring model, targeting greater than 98% precision. Flagged submissions still arrive — they’re just clearly marked, so nothing a real customer sends is silently dropped.

2. Summary and tagging

An AI-generated one-line summary and intent tag get attached automatically. A team member scanning the inbox can tell a wholesale inquiry from a returns request without opening either one.

3. Draft replies

Open the conversation, and Contactly can draft a reply grounded in what the customer actually asked — rewrite it for tone, or translate it, on request.

4. A human always sends

This is the part we won’t compromise on: AI drafts, you send. No reply reaches a customer without a person on your team reviewing it first. AI credits meter how much of this you use each month, scaled to your plan — unused credits don’t roll over.

The busywork — reading, tagging, drafting — gets handled automatically. The judgment calls stay with your team.

See the full breakdown on the Features page.