GCP Generative AI Leader Practice Question 82: Agents vs Applications Layer
Google Cloud Generative AI Leader · Test 9 · question 2 of 10
A company is comparing two gen AI initiatives. One team is building a custom procurement assistant that reasons across the purchasing, budgeting, and vendor systems, plans the steps needed, and completes a multi-step purchase approval on its own. A second team is adding a "summarize this expense report" button that calls Gemini inside the existing expense app to generate a short text summary for a person to read. In terms of the gen AI landscape's layers, how do these two initiatives differ?
Show answer and explanation
Answer:
The gen AI landscape is commonly described in layers: infrastructure, models, platforms, agents, and applications. The agents layer combines a foundation model with reasoning, memory, and tools so it can plan a multi-step task and take actions in the real environment, which matches a procurement assistant that reasons across systems and completes an approval on its own. The applications layer puts gen AI directly inside the products people already use to produce a single output, such as a summary a person reads and acts on themselves, which matches the expense button. Option B assigns each initiative to the correct layer.
- A is wrong because the models layer is the foundation model itself (for example Gemini), not what a team builds on top of it; collapsing both initiatives into that layer erases the real difference between planning multi-step action and generating a single response. - C is wrong because it swaps the two initiatives: a single button that returns a summary is the applications layer, while the multi-step, multi-system procurement flow is the agents layer. - D is wrong because the platforms layer refers to the tools used to build, customize, deploy, and manage a solution, not to what the finished solution does once it is running.
Reference: Generative AI Leader exam guide (section 1.3)