Email Assistant Setup Guide
The email assistant is our system that analyzes incoming emails with AI and automatically drafts replies or responds directly.
Step 1: Email Server Connection (IMAP & SMTP)
Go to the Email Accounts tab in your company panel and click the "Add New Account" button. Enter the required protocol details:
- Email address: The email that the assistant will read and reply to (e.g.,
support@yourcompany.com) - IMAP Server & Port: Use SSL/TLS supported Port 993 to read incoming emails (e.g.,
imap.gmail.comorimap.yandex.com). - SMTP Server & Port: Use TLS supported Port 587 or SSL supported Port 465 to send replies.
- Login Credentials: Write your email address and password. For security, your password is encrypted in our database using AES-256-GCM.
Critical Gmail / Outlook Setting (App Password)
For security reasons in Gmail or Microsoft 365 enterprise accounts, IMAP connection cannot be established with your normal password. You must go to your account security settings, enable 2-Step Verification, and then create an App Password. You must enter this 16-character special password in the password field.
Step 2: Running the Connection Test
Before saving the information, click the Test Connection button. The system must verify that it successfully logged in via both IMAP and SMTP. If the connection is successful, you can create the assistant.
Step 3: Draft Mode vs Live Mode
It is recommended to enable Draft Mode when first setting up the assistant:
- Draft Mode (Recommended): The AI writes the reply but does not send it directly to the customer. It saves it in the "Drafts" folder of your mailbox. You send it after checking and approving or editing it.
- Direct Send Mode: The AI prepares the response and sends it directly to the customer as an email. There is no human intervention in this mode.
Step 4: Working Hours and Holiday Mode
You can choose the hours (e.g., 09:00 - 18:00) or days your assistant will be active. For off-hours or holidays, you can configure the assistant to give static template replies without running the AI model. This prevents unnecessary credit consumption.