Lesson 1 — Where Analytics & AI/ML/DS Fit in Business
Why this matters¶
Many “AI projects” fail for simple reasons: unclear questions, weak data, and over-confident interpretation. This course treats AI as part of an evidence workflow, not a shortcut.
What is AI, Gen AI?¶
Artificial Intelligence (AI): Systems that can perform tasks that normally require human intelligence—such as recognizing patterns, making predictions, or recommending actions—by processing inputs (data, text, images) to produce outputs that support decisions.
Generative AI: A type of AI that can generate new content (text, code, images, etc.).
In this course, AI is treated as a tool within an evidence workflow (data → analysis → evaluation → communication) and will be used mainly as an assistant for tasks like planning an analysis, drafting explanations, and scaffolding code. Outputs must be verified (run the code, check the data, confirm sources) and should not be copied into assignments without understanding and proper disclosure where required.
The big picture: Analytics vs DS vs AI vs ML¶
A simple map (think tasks)¶
Describe: What happened? (EDA, visualization)
Explain: Why might it have happened? (regression, causal thinking)
Predict: What will happen next? (supervised ML)
Segment: Are there different types of users/markets? (unsupervised ML)
Later we’ll cover two learning modes: prediction with labels and pattern-finding without labels
Mini example: a churn workflow (end-to-end)¶
Question: Who is likely to churn next month?
Data: usage, plan type, support tickets, past churn.
EDA: churn rates by segment; distribution plots.
Model: start simple (logistic regression), then consider trees if needed.
Interpretation: what features matter? what are the limitations?
Decision: targeted retention + measurement plan.
Mini-lab (Google Colab)¶
Churn Data: https://
Review questions (quiz / reflection)¶
Give one example each of analytics, data science, and ML in business.
What is the difference between supervised and unsupervised learning?
Why can a strong model still lead to bad decisions?