Connecting the Dots

In the previous four lessons, we looked at the "Soloists"—Rekognition, Comprehend, audio tools, and Textract. But real-world business problems are rarely solved by a single service. On the AWS Certified AI Practitioner exam, the "Scenario" questions will test your ability to build a pipeline of services.

In our final lesson of Module 5, we will explore five "Archetype Scenarios" that appear frequently on the exam.

---

Scenario 1: The "Digital Customer Agent" (IVR)

Problem: A customer calls a support line. The company wants to understand who is calling, why they are calling, and how they feel.

The Pipeline:

1. Amazon Connect: Handles the telephony.
2. Amazon Transcribe: Converts the live audio to text.
3. Amazon Comprehend: Identifies the "Sentiment" (Angry?) and the "Intent" (Billing?).
4. Amazon Polly: Speaks the response back to the user.

---

Scenario 2: The "Intelligent Document Portal"

Problem: An insurance firm receives thousands of photos of car accidents and scanned claim forms. They want to automate the approval process.

The Pipeline:

1. Amazon Rekognition: Analyzes the photos to verify the damage (Object detection) and ensures no offensive content is present (Moderation).
2. Amazon Textract: Reads the claim form to pull out the Name, Policy Number, and repair estimate (Table/Form extraction).
3. Amazon Comprehend: Redacts the patient's PII (SSN, Phone) for privacy compliance before the adjuster sees it.

---

Scenario 3: The "Accessibility Engine"

Problem: A university wants to make its online video lectures accessible to students who are deaf and students who are blind.

The Pipeline:

1. Amazon Transcribe: Generates time-stamped subtitles for the deaf students.
2. Amazon Translate: Translates those subtitles into Spanish and French for international students.
3. Amazon Polly: Reads the lecture transcripts aloud for the visually impaired students.

---

Scenario 4: The "Market Intelligence" Bot

Problem: A marketing firm wants to monitor a "Live Stream" of a product launch on social media and detect whenever a specific influencer's face appears and what people are saying in the comments.

The Pipeline:

1. Amazon Rekognition Video: Detects the influencer (Celebrity Recognition) in the live video feed.
2. Amazon Comprehend: Analyzes the "Vibe" of the real-time chat comments to see if people are excited or bored.

---

Scenario 5: Global Content Moderation

Problem: A global social media startup needs to identify "Hate Speech" in 20 different languages and "Offensive Images" in user posts.

The Pipeline:

1. Amazon Rekognition: Filters the images for inappropriate content.
2. Amazon Translate: Converts all posts into English first.
3. Amazon Comprehend: Analyzes the translated text for "Toxicity" or "Negative Sentiment."

---

6. The "Quick-Reference" Scenario Table

Problem	Key Indicator	Primary Service
"Make text accessible as a voice"	Visual Impairment	Polly
"Find out why people are mad"	Sentiment Analysis	Comprehend
"Read an ink-on-paper form"	Structure/Handwriting	Textract
"Find landmarks/celeb in video"	Visual ID	Rekognition
"Translate text automatically"	Cross-language	Translate

graph TD
    A[Scenario Question] --> B{Is it Vision?}
    B -->|Yes| C[Rekognition / Textract]
    B -->|No| D{Is it Audio?}
    D -->|Yes| E[Transcribe / Polly]
    D -->|No| F{Is it Text Meaning?}
    F -->|Yes| G[Comprehend / Translate]

---

Summary of Module 5

We have mastered the Managed AI kingdom.

• You know how to see (Rekognition).
• You know how to read (Textract).
• You know how to extract meaning (Comprehend).
• You know how to hear and speak (Transcribe/Polly).

---

Knowledge Check

---

What's Next?

Managed services are powerful, but the future is Generative. In our next module, we look at the crown jewel of AWS AI: Module 6: Amazon Bedrock and Generative AI on AWS.