1.2.1. Recognizing Value and Appropriateness of AI/ML

First Principle: AI/ML is most valuable for problems that involve prediction, pattern recognition, or automation at a scale or complexity that is beyond human capability. It is not appropriate for tasks requiring 100% deterministic, rule-based outcomes.

Before starting an AI project, you must determine if it's the right tool for the job.

AI/ML provides value by:

• Assisting Human Decision Making: Flagging suspicious transactions for a fraud analyst to review.
• Solution Scalability: Categorizing millions of products on an e-commerce site, a task too large for a manual team.
• Automation: Automating repetitive tasks like reading data from forms.
• Personalization: Providing unique recommendations to millions of users.

AI/ML is NOT appropriate when:

• A simple, deterministic outcome is needed: If A + B must always equal C, you use traditional code, not a predictive model.
• There is no data: ML models cannot learn without data.
• The cost-benefit analysis is negative: The cost of developing and maintaining the model outweighs the business value it provides.
• The problem is not well-defined: You cannot build a model to predict an outcome if that outcome is not clearly defined and measurable.

Scenario: A manager suggests two projects: 1) An AI to calculate sales tax on an invoice. 2) An AI to predict which sales leads are most likely to close.

Reflection Question: Why is the first project a poor use of AI, while the second is an excellent one? Use the principles of determinism vs. prediction in your answer.

💡 Tip: Ask: "Is this a problem of calculation or prediction?" Calculation problems are for traditional software. Prediction problems are for AI/ML.