AI Agent Layer

Generating Smart, Context-Aware Support Responses

Overview

Once the message has been classified and routed, the AI Support Agent takes over. This is where GPT-4o (or a similar LLM) dynamically generates a structured, helpful, and brand-aligned response — drawing from past data, order history, FAQs, and your custom instructions.

This layer isn't just about replying — it’s about doing so intelligently:

Pulling in recent order info
Handling refunds, shipping updates, or common questions
Logging tickets when needed
Notifying your team when things go off script

1. Model Configuration

Parameter	Value
Model	OpenAI GPT-4o
Max Tokens	`1000–1500`
Temperature	`0.3–0.5` (balanced precision + tone)
JSON Mode	Enabled if you want structured actions
Memory/Context	Enabled (if using Qdrant or Pinecone)

2. System Prompt (Support Agent)

You’ll paste this prompt into the System Message section of the OpenAI node:

You are an AI customer support assistant.

You reply to messages on behalf of a company, using a helpful, calm, and respectful tone.

You may receive:

Frustrated messages

Questions about orders

Requests for refunds or help

General inquiries

Based on provided context (order info, company policies, FAQs), respond with:

A short, professional message to the customer

A flag if the issue should be escalated

A summary of what action was taken

If you don’t have enough info, ask a clarifying question or suggest a next step.

3. Input Data Passed to GPT-4o

Your input payload should include:

Cleaned customer message