Google Generative AI Leader Course Notes

Generative AI creates new content from patterns learned during training. On the exam, text, image, code, audio, and video generation usually point to foundation models rather than traditional analytics, because the model produces novel output instead of only classifying or forecasting an existing record.

• Confusing generative AI with classification or prediction; generative AI creates new content, while predictive systems estimate labels or outcomes.

• If the question says create new text, images, code, audio, or video, think generative AI and foundation models.

Vertex AI Studio is the quickest place to test prompts, compare model behavior, and refine examples before building a full application. It is best matched to prototyping because a team can experiment with Gemini settings and prompt patterns without first building a custom interface.

• Using Gemini for Workspace for a custom enterprise application when the scenario calls for Vertex AI, Agent Builder, or Enterprise Search.

• If the question says prototype prompts quickly, choose Vertex AI Studio.

Prompt engineering is the practice of shaping instructions so the model has enough goal, context, constraints, and output format to respond usefully. The exam often tests whether a vague request should be improved with clearer instructions rather than solved by changing the model itself.

• Trying to fix a vague prompt by changing the model when the best first step is clearer instructions, examples, or output format.

• If the question asks for examples of the desired answer pattern, choose few-shot prompting.

Generative AI strategy starts with business prioritization. Choose use cases where the problem is clear, users can explain what a good answer looks like, data is available, and the benefit can be measured with KPIs such as time saved, response quality, conversion rate, or cost reduction.

• Starting with technology selection before defining the business problem, KPI, risk, and data readiness.

• If the question asks how to prioritize use cases, look for business value, measurable KPIs, feasible data access, and manageable risk.