The Front Line of AI

For most people, their first interaction with a company's AI is via Customer Support. In the old world, this meant a frustrating phone tree. In the AWS AI world, this means an Intelligent Virtual Agent that can understand intent, sentiment, and specialized requests.

On the AWS Certified AI Practitioner exam, you will see scenarios about "Automating the Help Desk" or "Scaling Support."

---

1. The Core Service: Amazon Lex

Amazon Lex is a fully managed service for building conversational interfaces (chatbots) into any application using voice and text.

Wait, didn't we learn about Lex already?

• Amazon Polly: Turns text into a voice.
• Amazon Transcribe: Turns a voice into text.
• Amazon Lex: The "Brain" that manages the flow of the conversation. It creates the Chatbot.

Lex uses the same deep learning technology that powers Amazon Alexa.

---

2. Anatomy of a Lex Interaction

To build a bot in Lex, you need to understand three terms:

1. Intent: The goal the user wants to achieve. (e.g., "I want to book a hotel").
2. Utterance: The specific phrasing the user uses. (e.g., "I'd like to reserve a room" or "Book me a stay").
3. Slot: The specific pieces of data needed to complete the intent. (e.g., "Check-in date", "Room type").

---

3. The "Intelligent Contact Center": Amazon Connect

Amazon Connect is a cloud-based contact center (a virtual phone system). When you combine it with Lex, you get an AI-powered Phone System.

Business Application:

• Self-Service: The bot answers the phone, checks the customer's status in a database, and solves their problem without ever talking to a human.
• Sentiment-Based Routing: If Amazon Comprehend detects that a customer is "Angry" in the live chat, the system can bypass the bot and send the call to a "Retention Specialist" immediately.

---

4. Visualizing the Support Flow

graph TD
    A[Customer: 'Where is my order?'] --> B[Voice/Chat Interface]
    B --> C[Amazon Lex: Understanding Intent]
    C --> D{Is the data missing?}
    D -->|Yes| E[Lex: 'What is your order ID?']
    D -->|No| F[AWS Lambda: Search Database]
    F --> G[Order Found: 'Out for Delivery']
    G --> H[Amazon Polly: Voice Response]
    H --> I[Customer: 'Thank you!']
    
    subgraph Monitoring
    C --> J[Amazon Comprehend: Sentiment Check]
    J -->|Negative| K[Escalate to Human]
    end

---

5. Summary: High-Scale, Low-Cost Support

By using Lex and Connect, businesses can handle 10,000 calls simultaneously for a fraction of the cost of a human call center.

• The Practitioner's Goal: To automate the "Repetitive" questions (Where is my order? Reset my password) so that humans can focus on the "Complex" empathic problems.

---

Exercise: Identify the Bot Term

A user says: "I want to fly from New York to London on Tuesday." In Amazon Lex, what category does "New York" and "London" fall into?

• A. Intent.
• B. Utterance.
• C. Slot.
• D. Diarization.

The Answer is C! "New York" and "London" are specific data values (Slots) needed to fulfill the "Book Flight" intent.

---

Knowledge Check

What's Next?

Support is about reacting. What about "Acting"? In the next lesson, we see how to use AI to Personalize the world for your customers. Find out in Lesson 2: AI for Content Generation and Personalization.