You're building a descriptive analysis of appointment no-shows for a veterinary clinic in Nairobi — how bad the problem is, what the patterns look like, and whether it is getting worse.
The discipline skills: data profiling with pandas, computing rates and confidence intervals, running hypothesis tests with scipy, and producing a findings report a non-technical client can act on.
The AI-direction lesson: this is your first time directing AI through an analytical workflow. AI will produce confident, well-formatted output — summary statistics, charts, narrative. Some of it will be wrong in ways that are not obvious until you check. Wrong denominators in rate calculations. Misinterpreted p-values. The skill is checking AI's output against provided verification targets and catching the gap between "looks right" and "is right."
Your Role
You direct the analysis. AI does the computing, the plotting, the drafting. Your job is to tell it what to do, check what it produces, and make sure the numbers are right before they reach the client.
You have everything you need to verify the work: an analysis specification, verification targets, a data dictionary. The question is whether you use them.
What's New
This is the first real project. Everything is new: working with a client, directing AI through an analytical workflow, checking output against expected values, and producing a deliverable someone will actually read.
You have materials that tell you what to compute and what the results should look like. The challenge is not figuring out what to do. It is directing AI through the work and verifying that what comes back is accurate. AI produces confident, well-formatted output. Some of it will be wrong in ways that are not obvious until you check.
Tools
- Python 3.11+ via your conda "ds" environment
- Jupyter Notebook for the analysis
- pandas for data handling
- matplotlib / seaborn for visualization
- scipy for statistical tests
- Claude Code as the AI you direct
- Git / GitHub for version control
Materials
Everything is provided. You receive:
- A dataset of approximately 8,000 appointment records
- A data dictionary describing every column
- An analysis specification that defines what to compute
- Verification targets to check AI's output against
- A report template for the final deliverable
- Suggested prompts to get you started
Nothing requires prior statistical knowledge. The materials give you what you need to direct the work and verify the results.