FirstReply
FirstReply is a shared inbox for customer conversations with an AI assistant built into it. Messages from email, your website chat, WhatsApp, social channels and the app stores land in one place, the assistant classifies them, drafts or sends an answer from your own knowledge base, and hands anything it cannot answer to your team.
This documentation describes the product as a support team actually uses it: the inbox first, then the channels that feed it, then the AI and the knowledge base behind the answers.
Start here
First steps
Create your workspace, connect a channel and send the first reply.
The inbox
How conversations, assignment, statuses and replying work day to day.
How the AI works
Classification, knowledge retrieval, automatic replies and hand-over.
Knowledge base
The source of every answer the assistant gives. Sync it or write it.
What FirstReply does
One inbox for every channel. An email to support@, a message in your
website chat widget and a WhatsApp question all become conversations in the
same list, with the same assignment, statuses and search.
Answers from your knowledge base, not from guesswork. The assistant retrieves passages from the articles you synced or wrote, and answers from those. When the knowledge base does not cover a question, it says so and hands the conversation over instead of inventing an answer.
Classification that routes work. Every incoming message is sorted into one of your categories. Categories decide whether the AI replies by itself, drafts for review, assigns to a specific person, or stays fully manual.
A queue that surfaces what matters. The AI Queue collects the conversations that need a human now: urgent ones, unhappy customers, messages waiting more than a day, and drafts ready for review.
How the pieces fit together
A message arrives on a channel, becomes a conversation, gets classified, and from there follows the rules of its category. Everything the assistant does stays visible in the conversation, and your team can take over at any point.
Conventions in this documentation
The screenshots show a demo workspace for Lumo Bikes, a fictional e-bike brand. Its customers, orders and knowledge base articles are invented, so nothing you see is real customer data.
Where a feature depends on your role in the workspace, the requirement is stated in the text. Roles are explained under Members and roles.