Présentation du cours
AI can accelerate analysis, but it cannot rescue an unclear question or an untrustworthy table. This lab begins with the sheet itself: structure, definitions, missing values, and the checks that make a conclusion defensible.
What you will build
You will take a small operational dataset from raw rows to a one-page decision note. The work includes a data dictionary, a cleaning log, a validated calculation, an honest chart, and a reproducible handoff.
Inside the lab
- Reshape a spreadsheet into one row per observation with clear headers.
- Ask for an analysis plan before asking for an answer.
- Inspect formulas, generated code, filters, and denominators.
- Choose a chart that matches the decision—not the prettiest option.
- Protect sensitive data and neutralize spreadsheet-export risks.
Best for: intermediate learners working in ChatGPT, Excel with Copilot, or similar analysis tools. Python and Copilot availability depend on license and update channel.
Last reviewed September 2026. Lessons cite current OpenAI, Microsoft, NIST, and OWASP documentation.