Where is the Money?

As a leader, your job isn't just to deploy AI; it's to deploy AI that matters. "We built a chatbot" is not a business result. "We reduced call center handle time by 30%" is a business result.

To find these results, researchers and efficient leaders categorize Generative AI use cases into three specific buckets: Creation, Summarization, and Discovery.

In this lesson, we break down these categories to help you spot opportunities in your own organization.

---

1. Creation (The "Writer")

This is the most obvious capability: Generating novel content.

• Mechanism: Text-to-Text, Text-to-Image, Text-to-Code.
• Business Goal: Increase productivity of creative/technical workers.

Examples:

• Marketing: Generating 50 variations of an Instagram ad copy to A/B test.
• Engineering: Providing boilerplate code or unit tests (using Gemini Code Assist).
• Sales: Drafting personalized outbound emails to 1,000 prospects.

The Exam Trap: Do not confuse "Creation" with "Accuracy." Creation is best when there is no single right answer.

• Good: "Write a funny poem."
• Bad: "Write the chemical formula for water." (This is retrieval, not creation).

---

2. Summarization (The "Reader")

This is the huge efficiency unlock. LLMs are incredible at compression. They can take a massive amount of unstructured data and squeeze it into a structured insight.

• Mechanism: Long-Context Input -> Concise Output.
• Business Goal: Reduce cognitive load and manual processing time.

Examples:

• Customer Support: Summarizing a 20-minute audio call transcript into 3 bullet points: "Issue, Resolution, Sentiment."
• Legal: Scanning a 50-page contract to extract "Termination Clauses."
• Meetings: Using Gemini for Workspace to summarize a Google Meet recording.

Value Metric: Time Saved per Employee. If legal team spends 5 hours reading a contract, and AI does it in 10 seconds, the ROI is massive.

---

3. Discovery (The "Librarian")

This is the ability to find answers and insights within complex data using natural language. It powers the RAG and Search applications we discussed in Module 2.

• Mechanism: User Question -> Search -> Answer.
• Business Goal: Access to knowledge and faster decision making.

Examples:

• HR: "How do I add a newborn to my dental plan?" (The AI finds the PDF and explains the steps).
• Maintenance: "Show me all reports where the hydraulic pump failed due to overheating." (The AI searches unstructured logs).
• Strategy: "What were the top 3 complaints from customers in Q4?"

---

4. Visualizing the Value Matrix

When brainstorming with your team, map ideas to this chart:

graph TD
    Start{What does the Human need?}
    
    Start -->|New Content?| Creation
    Start -->|Less Noise?| Summarization
    Start -->|An Answer?| Discovery
    
    subgraph "Creation: Speed & Variety"
    Creation --> Mkt[Marketing Copy]
    Creation --> Code[Code Assist]
    end
    
    subgraph "Summarization: Efficiency"
    Summarization --> Docs[Document Processing]
    Summarization --> Call[Call Analysis]
    end
    
    subgraph "Discovery: Knowledge"
    Discovery --> Search[Enterprise Search]
    Discovery --> Insights[Data Q&A]
    end
    
    style Creation fill:#F4B400,stroke:#fff,stroke-width:2px,color:#fff
    style Summarization fill:#4285F4,stroke:#fff,stroke-width:2px,color:#fff
    style Discovery fill:#34A853,stroke:#fff,stroke-width:2px,color:#fff

---

5. Code Example: Which API Call is Which?

Technically, these all use the same generate_content method, but the prompt structure defines the category.

# 1. CREATION Prompt
prompt_create = "Write a Python function to calculate Fibonacci numbers."

# 2. SUMMARIZATION Prompt
prompt_summary = """
Read the following 5000 words of text.
Provide a 3-sentence summary of the main arguments.
TEXT: [Insert Text]
"""

# 3. DISCOVERY Prompt (requires Grounding/RAG)
prompt_discovery = """
Based on the provided Knowledge Base, answer:
'What is the procedure for a fire alarm drill?'
"""

---

6. Summary

• Creation: Making new things (Code, Images, Text). Value = Speed.
• Summarization: Compressing big things. Value = Efficiency.
• Discovery: Finding hidden things. Value = Knowledge.

Leadership Tip: The highest ROI usually comes from Summarization and Discovery first. Why? Because they solve "drudgery" (reading boring docs). Creation is impactful but risky because it requires high human quality control.

In the next lesson, we will look at Agents, the advanced evolution of these concepts where AI doesn't just "talk," but actually "does work."

---

Knowledge Check