Proving the Worth

Machine Learning and Generative AI can be expensive. Between developer salaries, AWS compute costs (Training), and token costs (Inference), a project can easily cost $50,000 before it solves a single problem.

As an AWS AI Practitioner, you must be able to calculate the ROI (Return on Investment) to ensure you are creating more value than you are consuming.

---

1. The Simple ROI Formula

In its basic form: ROI = (Net Value Gained - Cost of Project) / Cost of Project * 100

Example: The Support Bot

• Annual Cost of Human Support: $200,000.
• Cost to Build AI Bot: $20,000 (One-time).
• Cost to Run AI Bot (AWS tokens): $10,000 / year.
• Efficiency Gain: Bot handles 50% of the load.
• Value Gained: $100,000 (50% of human salary saved).
• ROI (Year 1): $(100,000 - 30,000) / 30,000 = 233%$.

---

2. The Three Dimensions of Value

Value isn't always just "Money in the bank." AWS looks at three types of value:

Dimension A: Cost Reduction (Bottom Line)

• Reducing the time it takes to process a loan.
• Reducing electricity waste using a forecast model.
• Reducing fraud losses using Amazon Fraud Detector.

Dimension B: Revenue Growth (Top Line)

• Increasing the "Add to Cart" rate by 5% using Amazon Personalize.
• Opening a new market by translating your app into 10 new languages using Amazon Translate.

Dimension C: Intellectual Property & Agility (Intantigble)

• Becoming an "AI-First" company makes it easier to hire top talent.
• Being able to "Ask a question" of your data using Amazon Q allows you to make decisions in minutes instead of weeks.

---

3. CODN: The Cost of Doing Nothing

This is a powerful concept for the exam. Sometimes, the "Cost" of AI seems high, but the Cost of Doing Nothing is higher.

• If your competitor implements a GenAI support bot and can respond to customers 24/7, and you don't, you will lose market share.
• In this case, even if the "ROI" is low in year 1, the CODN (losing your business) makes the project a "Must-do."

---

4. Visualizing the Project Lifecycle Costs

graph TD
    subgraph Investment_Phase
    A[Data Selection/Cleaning]
    B[Model Training/Fine-tuning]
    C[Integration/App Dev]
    end
    
    subgraph Operational_Phase
    D[Token Costs / Instance Costs]
    E[Monitoring / Maintenance]
    F[Security Audits]
    end
    
    subgraph Value_Realization
    G[Time Saved]
    H[Errors Avoided]
    I[New Sales]
    end
    
    A & B & C --> D & E & F
    D & E & F --> G & H & I
    Note[ROI happens where Value > Operation + Investment]

---

5. Summary: Be Conservative

A good Practitioner always "Over-estimates costs" and "Under-estimates value" in their pitch.

• Hidden Cost Alert: Don't forget the cost of cleaning the data. This is often the most expensive part of a SageMaker project.

---

Exercise: Identify the Value Type

A logistics company uses Amazon Forecast to predict truck maintenance. By fixing trucks before they break down on the highway, they save $5,000 in towing fees and lost driver time per incident. What type of value is this?

• A. Revenue Growth.
• B. Intellectual Property.
• C. Cost Reduction.
• D. Data Security.

The Answer is C! You are reducing an existing, recurring expense (towing/lost time).

---

Knowledge Check

What's Next?

Calculating the math is easy; moving the people is hard. In our final lesson of Module 9, we look at the human side of AI. Find out in Lesson 4: Change Management for AI Adoption.