Understanding Intent, Urgency, and Path Forward

Overview

Now that a message has been received and preprocessed, the system uses Gemini 1.5 Flash to determine what the message is really about — and how it should be handled.

This layer analyzes:

• The emotional tone of the user
• The message category (e.g. refund, order issue, general support, feedback)
• The confidence score
• The channel it came through (email, Telegram, etc.)

These insights are used to make critical routing decisions:

Should this message be answered immediately by the AI?

Should it be escalated to a human agent?

Does it require creating a support ticket, updating an order, or logging feedback?

How It Works

1. Send Message to Gemini Flash

• Use an HTTP Request Node in n8n or a prebuilt Gemini Node.
• Input the cleaned/transcribed message as prompt or content.
• Use structured system instructions like:

You are a classification AI.
Determine the emotional tone and purpose of the following message.
Respond only in JSON format with:
- tone
- intent category
- source channel
- confidence level (0–1)

2. Example Output from Gemini