Skip to content

Chapter 6: Webex Contact Center Implementation -- 6.8 AI Platform Integration

6.8 AI Platform Integration

Why Two AI Platforms?

Abhavtech's Virtual Agent "Abhi" uses both Webex AI Agent and Google Dialogflow CX — not as alternatives, but as a hybrid architecture where each handles what it does best.

Platform Handles Why
Webex AI Agent Simple, single-turn intents — office hours, queue status, callback requests, language selection Native to WxCC, zero integration cost, fast to configure, sufficient for ~60% of call volume
Google Dialogflow CX Complex multi-turn dialogues — order tracking (needs order ID + account lookup), billing disputes (multi-step), product registration Full ML-powered NLU, stateful conversation, webhook fulfillment to Salesforce/billing DB, handles ambiguity

Decision rule: If a caller intent can be resolved in ≤ 2 exchanges with no backend DB lookup → Webex AI Agent. If it requires conversation state, entity extraction, or REST API calls → Dialogflow CX.

The CCAI Connector in WxCC routes each call to the appropriate engine based on the initial intent detected. See 10.1.3 Hybrid Architecture for the full design rationale and traffic flow split.

6.8.1 Architecture Overview

+-----------------------------------------------------------------------------+
|              VIRTUAL AGENT "ABHI" - DUAL PLATFORM ARCHITECTURE              |
+-----------------------------------------------------------------------------+
|                                                                             |
|  +---------------------------------------------------------------------+   |
|  |                     GOOGLE CLOUD PLATFORM                            |   |
|  |  +-------------+  +-------------+  +-------------+                  |   |
|  |  | Dialogflow  |  | Cloud       |  | Cloud       |                  |   |
|  |  | CX Agent    |  | Functions   |  | Speech/TTS  |                  |   |
|  |  | "Abhi"      |  | (Webhooks)  |  | APIs        |                  |   |
|  |  +------+------+  +------+------+  +-------------+                  |   |
|  |         |                |                                           |   |
|  |         +----------------+                                           |   |
|  |                |                                                     |   |
|  +----------------+-----------------------------------------------------+   |
|                   | Service Account + JSON Key                              |
|                   v                                                         |
|  +---------------------------------------------------------------------+   |
|  |                     WEBEX CONTACT CENTER                             |   |
|  |  +-------------+  +-------------+  +-------------+                  |   |
|  |  | Virtual     |  | Flow        |  | Agent       |                  |   |
|  |  | Agent       |  | Designer    |  | Desktop     |                  |   |
|  |  | Connector   |  | VA Node     |  | (Context)   |                  |   |
|  |  +-------------+  +-------------+  +-------------+                  |   |
|  |                                                                      |   |
|  +---------------------------------------------------------------------+   |
|                                                                             |
+-----------------------------------------------------------------------------+

6.8.2 PART A: Google Cloud Platform Configuration

6.8.2.1 Prerequisites

Requirement Details
Google Cloud Account With billing enabled
IAM Permissions Dialogflow API Admin, Service Account Admin
APIs to Enable Dialogflow, Speech-to-Text, Text-to-Speech
Region asia-south1 (Mumbai) for data residency

6.8.2.2 Step 1: Create Google Cloud Project

Navigation: console.cloud.google.com

  1. Click "Select Project" -> "New Project"
  2. Project Name: abhavtech-wxcc-ai
  3. Organization: abhavtech.com
  4. Click "Create"
  5. Record Project ID: abhavtech-wxcc-ai

6.8.2.3 Step 2: Enable Required APIs

Navigation: APIs & Services -> Library

Enable: - [OK] Dialogflow API - [OK] Cloud Speech-to-Text API - [OK] Cloud Text-to-Speech API - [OK] Cloud Functions API - [OK] Secret Manager API

6.8.2.4 Step 3: Create Service Account

Navigation: IAM & Admin -> Service Accounts

  1. Click "Create Service Account"
  2. Name: wxcc-dialogflow-connector
  3. Description: Service account for Webex Contact Center integration
  4. Click "Create and Continue"

Assign Roles: - Dialogflow API Client - Dialogflow API Reader - Service Account Token Creator

  1. Click "Done"

Generate JSON Key:

  1. Click on service account wxcc-dialogflow-connector
  2. Go to "Keys" tab
  3. Click "Add Key" -> "Create New Key"
  4. Select JSON
  5. Click "Create"
  6. SAVE the downloaded JSON file securely

6.8.2.5 Step 4: Create Dialogflow CX Agent

Navigation: dialogflow.cloud.google.com/cx

  1. Select Project: abhavtech-wxcc-ai
  2. Click "Create Agent"
  3. Configure:
+-----------------------------------------------------------------------------+
|  DIALOGFLOW CX AGENT CONFIGURATION                                         |
+-----------------------------------------------------------------------------+
|                                                                             |
|  FIELD                      | VALUE                                        |
|  ---------------------------+--------------------------------------------|
|  Display Name               | Abhi (अभि)                                   |
|  Location                   | asia-south1 (Mumbai)                         |
|  Default Language           | en (English)                                 |
|  Additional Languages       | hi (Hindi)                                   |
|  Time Zone                  | Asia/Kolkata                                 |
|  ---------------------------+--------------------------------------------|
|                                                                             |
|  ADVANCED SETTINGS:                                                        |
|  * Speech to Text: Google Cloud Speech                                     |
|  * Text to Speech: Neural voice (en-IN-Wavenet-A, hi-IN-Wavenet-A)        |
|  * Logging: Enable Cloud Logging                                           |
|  * Enable spell correction: Yes                                            |
|                                                                             |
+-----------------------------------------------------------------------------+
  1. Click "Create"
  2. Record Agent ID: (from Agent Settings)

6.8.2.6 Step 5: Configure Intents (Per Chapter 3.9.2)

Phase 1 Intents - 15 Intents:

+-----------------------------------------------------------------------------+
|              PHASE 1 INTENT LIBRARY - 15 INTENTS                            |
+-----------------------------------------------------------------------------+
|                                                                             |
|  INTENT               | TRAINING PHRASES | FULFILLMENT          | CHANNEL  |
|  ---------------------+------------------+----------------------+----------|
|  greeting.hello       | 20               | Greeting response    | All      |
|  order.status         | 30               | Webhook: /orders     | All      |
|  order.track          | 25               | Webhook: /tracking   | All      |
|  product.inquiry      | 40               | Knowledge Base       | All      |
|  product.pricing      | 25               | Webhook: /pricing    | All      |
|  account.balance      | 20               | Webhook: /balance    | Voice    |
|  account.info         | 20               | Webhook: /account    | All      |
|  support.general      | 35               | Escalate to agent    | All      |
|  support.troubleshoot | 30               | KB + escalate        | All      |
|  billing.inquiry      | 25               | Webhook: /billing    | Voice    |
|  hours.location       | 15               | Static response      | All      |
|  callback.request     | 15               | Callback flow        | Voice    |
|  agent.handoff        | 20               | Immediate escalate   | All      |
|  feedback.survey      | 15               | Survey flow          | All      |
|  fallback.default     | N/A              | Clarify + retry      | All      |
|  ---------------------+------------------+----------------------+----------|
|                                                                             |
|  TOTAL: 15 Intents, ~355 Training Phrases                                  |
|                                                                             |
+-----------------------------------------------------------------------------+

Intent Creation Procedure:

  1. Navigate to "Manage" -> "Intents"
  2. Click "+ Create"
  3. Enter Intent Name: order.status
  4. Add Training Phrases (30):
  5. Where is my order
  6. Track my order
  7. Order status
  8. What is my order status
  9. Mera order kahan hai (Hindi)
  10. Order ka status kya hai (Hindi)
  11. (Add remaining variations)
  12. Click "Save"

REPEAT for all 15 intents.

6.8.2.7 Step 6: Configure Webhook Fulfillment

Create Cloud Function:

Navigation: Cloud Functions -> Create Function

// abhavtech-va-webhook/index.js
const functions = require('@google-cloud/functions-framework');

functions.http('dialogflowWebhook', async (req, res) => {
  const tag = req.body.fulfillmentInfo?.tag;
  const params = req.body.sessionInfo?.parameters || {};
  const ani = params.ani || '';

  let responseText = '';

  switch(tag) {
    case 'order-status':
      const orderNum = params.order_number;
      // Call Abhavtech Order API
      const status = await getOrderStatus(orderNum);
      responseText = `Your order ${orderNum} is ${status.status}. 
                      Expected delivery: ${status.delivery_date}.`;
      break;

    case 'account-balance':
      const balance = await getAccountBalance(ani);
      responseText = `Your current account balance is ₹${balance}.`;
      break;

    case 'product-pricing':
      const product = params.product_name;
      const pricing = await getProductPricing(product);
      responseText = `${product} pricing starts at ₹${pricing.base_price}. 
                      Would you like to know more?`;
      break;

    default:
      responseText = 'I understand. Let me connect you with an agent.';
  }

  res.json({
    fulfillment_response: {
      messages: [{
        text: { text: [responseText] }
      }]
    }
  });
});

// API helper functions
async function getOrderStatus(orderNum) {
  const response = await fetch(`${process.env.ABHAVTECH_API}/orders/${orderNum}`);
  return response.json();
}

async function getAccountBalance(phone) {
  const response = await fetch(`${process.env.ABHAVTECH_API}/customers/balance?phone=${phone}`);
  return response.json();
}

async function getProductPricing(product) {
  const response = await fetch(`${process.env.ABHAVTECH_API}/products/pricing?name=${product}`);
  return response.json();
}

Deploy Function:

gcloud functions deploy abhavtech-va-webhook \
  --gen2 \
  --runtime=nodejs18 \
  --region=asia-south1 \
  --trigger-http \
  --allow-unauthenticated \
  --set-env-vars ABHAVTECH_API=https://api.abhavtech.com/v1 \
  --entry-point=dialogflowWebhook

Configure in Dialogflow CX:

  1. Navigate to "Manage" -> "Webhooks"
  2. Click "+ Create"
  3. Display Name: abhavtech-fulfillment-webhook
  4. Webhook URL: (Cloud Function URL from deployment)
  5. Click "Save"

6.8.2.8 Step 7: Configure Escalation Flow

  1. Create Intent: agent.handoff
  2. Training Phrases:
  3. Talk to an agent
  4. I want to speak to someone
  5. Human please
  6. Connect me to a person
  7. Agent
  8. Representative

  9. Mujhe agent se baat karni hai (Hindi)

  10. Create Escalation Page:

  11. Add Parameter: live_agent_escalation = true
  12. Response: "I'm connecting you to an agent now. Please hold."

6.8.2.9 Step 8: Test Agent

  1. Click "Test Agent" in Dialogflow CX Console
  2. Test each intent with sample utterances
  3. Verify webhook responses
  4. Test Hindi language
  5. Test escalation flow

6.8.3 PART B: Webex Contact Center Configuration

6.8.3.1 Step 1: Add Virtual Agent Connector

Navigation: Control Hub -> Contact Center -> Connectors

  1. Click "Set Up" under Virtual Agent
  2. Select "Google Dialogflow CX"
  3. Click "Next"

6.8.3.2 Step 2: Configure Connection

+-----------------------------------------------------------------------------+
|  WXCC DIALOGFLOW CX CONNECTOR CONFIGURATION                                |
+-----------------------------------------------------------------------------+
|                                                                             |
|  FIELD                      | VALUE                                        |
|  ---------------------------+--------------------------------------------|
|  Connector Name             | Abhi-Dialogflow-Connector                    |
|  Google Cloud Project ID    | abhavtech-wxcc-ai                            |
|  Location                   | asia-south1                                  |
|  Agent ID                   | [From Dialogflow CX Agent Settings]          |
|  Service Account Key        | [Upload JSON key file]                       |
|  ---------------------------+--------------------------------------------|
+-----------------------------------------------------------------------------+
  1. Upload Service Account JSON key
  2. Click "Test Connection" - Verify "Connection Successful"
  3. Click "Save"

6.8.3.3 Step 3: Create Virtual Agent Configuration

Navigation: Control Hub -> Contact Center -> Features -> Virtual Agent

  1. Click "Create Virtual Agent"
  2. Configure:
Field Value
Name Abhi
Connector Abhi-Dialogflow-Connector
Default Language en-US
Additional Languages hi-IN
Voice Neural (Dialogflow TTS)
Max No Input 3
Max Turns 10
Timeout Per Turn 30 seconds
  1. Click "Save"

6.8.3.4 Step 4: Add Virtual Agent Node to Flow

Navigation: Contact Center -> Flows -> India_MainMenu_Flow_v1 -> Edit

  1. Add "Virtual Agent V2" node after language selection
  2. Configure:
+-----------------------------------------------------------------------------+
|  VIRTUAL AGENT V2 NODE - FLOW DESIGNER CONFIGURATION                       |
+-----------------------------------------------------------------------------+
|                                                                             |
|  FIELD                      | VALUE                                        |
|  ---------------------------+--------------------------------------------|
|  Virtual Agent              | Abhi                                         |
|  Input Audio                | {{NewPhoneContact.MediaResourceId}}          |
|  Language                   | {{selected_language}}                        |
|  Welcome Event              | WELCOME                                      |
|  Timeout Per Turn           | 30 seconds                                   |
|  Max No Input               | 3                                            |
|  ---------------------------+--------------------------------------------|
|                                                                             |
|  OUTPUT PATHS:                                                             |
|  * Handled -> End Flow (contained call)                                     |
|  * Escalated -> Continue to DTMF Menu (Node 8)                             |
|  * Error -> Error handling -> Queue_Support                                  |
|                                                                             |
|  CONTEXT VARIABLES TO PASS TO AGENT:                                       |
|  * va_summary = {{VirtualAgent.TranscriptSummary}}                        |
|  * va_intent = {{VirtualAgent.LastIntent}}                                |
|  * va_reason = {{VirtualAgent.EscalationReason}}                          |
|                                                                             |
+-----------------------------------------------------------------------------+
  1. Connect output paths
  2. Validate and Publish flow

6.8.3.5 Virtual Agent Testing Checklist

+-----------------------------------------------------------------------------+
|              VIRTUAL AGENT "ABHI" - TESTING CHECKLIST                       |
+-----------------------------------------------------------------------------+
|                                                                             |
|  GOOGLE CLOUD SIDE:                                                        |
|  [ ] Service Account permissions correct                                   |
|  [ ] Dialogflow agent responds in test console                            |
|  [ ] Webhook integration works (order status, balance)                    |
|  [ ] Hindi language model trained and working                             |
|  [ ] Escalation intent triggers properly                                  |
|                                                                             |
|  WEBEX SIDE:                                                               |
|  [ ] Connector test passes ("Connection Successful")                      |
|  [ ] Virtual Agent node executes in flow                                  |
|  [ ] Speech recognized correctly                                          |
|  [ ] Contained calls end properly                                         |
|  [ ] Escalated calls route to queue with context                          |
|  [ ] Context (summary, intent, reason) visible in Agent Desktop           |
|                                                                             |
|  END-TO-END SCENARIOS:                                                     |
|  [ ] "Where is my order" -> Order status provided -> Call ends              |
|  [ ] "What is my balance" -> Balance stated -> Call ends                    |
|  [ ] "Talk to agent" -> Escalates with context -> Agent receives call       |
|  [ ] Hindi: "Mera order kahan hai" -> Status in Hindi -> Call ends          |
|  [ ] 3x no input -> Auto escalate -> Agent receives call                    |
|                                                                             |
|  CONTAINMENT TARGET: 25% (Phase 1)                                         |
|                                                                             |
+-----------------------------------------------------------------------------+