Intercom Conversation Reviews

# 🧠 Intercom Conversation QA Review Automation – n8n Workflow

## ✅ Objective

Automate the QA review of Intercom support conversations using AI to:

* Evaluate agent performance
* Log structured feedback
* Provide coaching where needed
* Track trends in customer interactions

---

## 🧩 Workflow Overview

### 1. **Webhook Trigger**

* Listens for `conversation.admin.closed` events from Intercom.
* Initiates the workflow whenever a conversation is closed.

### 2. **Fetch Conversation Details**

* Uses Intercom API to retrieve the full conversation using `conversation ID`.

### 3. **Filter Irrelevant Conversations**

* Skips conversations marked as “Spam/Promotional” or “Other” via `custom_attributes`.

### 4. **Structure the Conversation**

* Extracts:

  * Message body
  * Role (User, Bot, Admin)
  * Author name
  * Timestamps
* Formats messages into a readable transcript.

### 5. **AI-Powered Evaluation (Parallel Branches)**

#### A. 🧪 **Performance Scoring (GPT-4o)**

Evaluates the following attributes:

* **Response Time**
* **Clarity of Answers**
* **Tone & Behavior**
* **Urgency Handling**
* **Ownership & Accountability**
* **Problem Solving**
* Outputs:

  * `Start Date`, `End Date`
  * `Overall Score` (1 to 5)
  * Attribute-wise scores

#### B. 🔍 **Message Analytics**

* Counts the number of messages from:

  * Bot
  * User
  * Admin
* Identifies the **last admin responder's name**

### 6. **Merge & Log**

* Combines scoring and metadata
* Appends the data to a Google Sheet:

  * **Sheet:** `Daily Input Update`
  * **Tab:** `AI.Coach`

### 7. **Optional: Coaching Feedback Generator (Score ≤ 3)**

* Triggered if the score is 3 or below
* Generates coaching feedback in this format:

  * ✅ What went well
  * ❌ What needs fixing (with quote + suggestion)
  * 😕 How the customer likely felt
  * 🔄 What to do differently next time

---

## 📈 Example Output

| Conversation ID | Response Time | Clarity | Tone   | Ownership | Overall Score | Last Message By |
| --------------- | ------------- | ------- | ------ | --------- | ------------- | --------------- |
| 215469787223351 | Fast          | Clear   | Polite | Strong    | 5             | Kinjal          |

---

## 💡 Why It Matters

* Automates repetitive QA work at scale
* Transforms unstructured conversations into actionable insights
* Boosts support quality, ownership, and customer satisfaction
* Offers personalized, high-signal feedback for team coaching

---