Skip to content

🧪 Practice Lab: AI Strategist

Welcome to the executive boardroom of MagicFridge. Here, we no longer talk about code, but about governance, budget, and people. Your mission is to deploy AI without putting the company in danger.

Exercise 1: Tracking Shadow AI 👻

(Objective: identify shadow AI risks - LO 5.1.1)

Situation: during an audit, you discover that a junior developer copied the entire source code of the MagicFridge "Bank Payment" module into ChatGenius (a free, public AI tool) to ask it for optimizations.

What is the major immediate risk?

Your diagnosis

●
●
●

Exercise 2: Choosing the Right Weapon (LLM vs. SLM) ⚔️

(Objective: select the appropriate model - LO 5.1.3)

Project: MagicFridge wants to launch an "Offline Assistant" feature: the user must be able to ask for a recipe in natural language even at the back of a supermarket without an internet connection (the AI runs directly on the phone).

Which type of model do you choose?

The architect's choice


Exercise 3: Job Evolution 💼

(Objective: understand new skills - LO 5.2.1)

Context: you are writing the job description to recruit a new "GenAI QA Tester". Among these three skills, which one is specific to this new role (and not classic testing)?

  1. Knowing how to write test cases in Gherkin.
  2. Knowing how to evaluate model temperature and context window.
  3. Knowing how to use Jira for bug tracking.
See the answer

Answer: 2. Knowing how to evaluate model temperature and context window.

  • Why? This is a technical skill purely related to how LLMs function.
  • Skills 1 and 3 are classic software testing skills (and still necessary!), but not specific to GenAI.

Exercise 4: The Deployment Dilemma 🚀

(Objective: manage change and risks - LO 5.1.4)

Situation: management is impatient. The CEO wants to deploy GUS everywhere, immediately, to replace customer service by next Monday. The QA team has only done a few promising exploratory tests (Discovery Phase).

What is your recommendation as a Test Manager?

  1. "Let's go! The AI will learn on the job with real customers."
  2. "Stop. We must go through the Initiation and Usage Definition phase before generalizing."
See the recommendation

Answer: 2. Stop.

Skipping steps is a recipe for disaster (massive hallucinations, angry customers). We must first:

  1. Select a pilot use case.
  2. Set up guardrails (RAG, filters).
  3. Measure quality before opening the floodgates.


Training Complete!

🎉 Congratulations! You have covered the entire curriculum.

If this free course helped you feel ready for your certification exam, consider supporting the guide: