Re:amaze vs Re:amaze

Re:amaze publishes a blog that doubles as its release channel, and the run since January is a single continuous build: the AI Agent launched as a knowledge-base-backed chatbot, gained smarter intent detection, extended to email and SMS, then a mid-year roundup added visibility into how the agent works and ways to test it before it goes live. The newest entry introduces an AI Pre-Filter that sorts the inbox before any of that runs — separating marketing mail, shipping notices, social mentions and spam from the customers who actually need an answer.

The company is working backwards from the reply toward the inbox. First the agent answered questions, then it understood them better, then it reached more channels, then it became inspectable and testable. Pre-filtering is the logical next step upstream: deciding what is even a support conversation. The supporting content — how to write help centre articles for better AI answers, how much support AI should handle — is aimed at customers who have to decide how much of the queue to hand over.

Expect the pre-filter and the agent to be connected into a single routing decision, with escalation rules for the VIP case the announcement explicitly calls out as needing a person.