Week 1 — Environment & Language Basics · ~15 min
Using AI in this course: the rules
Free previewWhy this lesson comes first
ChatGPT, Claude, Gemini and Copilot can write a pandas script in seconds. Ask for "revenue by region from this CSV" and you'll get code that runs. So why learn to write it yourself?
Because in analysis, code that runs is not the same as an answer that's right. A merge that silently doubles rows, a filter that drops every blank, an average over the wrong population: all of these run without an error and produce a confident, wrong number. Someone then makes a decision with it. Your job is to be the person who can tell.
You can only check an analysis you could have done yourself. So this course brings AI in gradually.
The three stages
| Stage | Weeks | What AI may do | What you must do |
|---|---|---|---|
| Foundations | 1–6 | Explain code, concepts and error messages, like a patient tutor. | Type every line of your own code. |
| Pair | 7–11 | Suggest approaches, draft functions and pandas code, generate fake test data. | Run it, check it against known totals, and explain every line in a spot check. |
| Lead | 12 (final project) | Anything a teammate could do. | Own the result: the checks, the summary, and a live defence of every number. |
The rule underneath all three: AI may explain; code you can't walk through line by line is not your work.
Rule 1: Disclose your AI use
Every submission includes an AI-use note. There's a box for it on the submission form:
AI use: asked Claude why my merge gave 10 rows instead of 8
What I checked: printed shapes before/after; found Bola twice in customers.csv
What it got wrong: suggested drop_duplicates() on the merged result, which hid the problem
"No AI used" is a fine answer. Honest disclosure never costs marks. Hidden AI use that we discover in a spot check is treated as not your work.
Rule 2: Never paste these into an AI tool
- Real people's data: customer lists, sales records with names or phone numbers, staff records, anything with BVN, NIN, bank or health details. Not even "just the first 10 rows".
- An employer's or client's data or code, unless their written policy allows it.
- Passwords, API keys, tokens, and the contents of
.envfiles.
Under the Nigeria Data Protection Act 2023 (NDPA), personal data may only be processed for a lawful purpose, kept secure, and shared only with appropriate safeguards. Pasting it into a public chatbot sends it to a third party you don't control. When you need to show an AI your data's shape, describe the columns or invent a few fake rows with the same structure. That's a skill in itself.
Rule 3: Verify against known totals
AI is fluent and often wrong. In data work that looks like:
- functions and arguments that don't exist (
df.groupby("region").total(),pd.read_xlsx); - code that answers a slightly different question from the one you asked;
- merges, filters and averages that run cleanly and give the wrong number.
The defence is the habit this course drills from Week 1: known input, known output. The practice data has 8 rows, revenue 160, Efe has no orders, and monthly revenue is 35 / 65 / 60. Any AI-written code must reproduce those before you trust it on anything else.
Rule 4: You own the numbers
When your summary says "North drives 41% of revenue", you're the one answering for it, not the tool. In the final project you'll trace numbers back to their cells live. Use AI to go faster, never to skip understanding.
Sign in to track your progress.