GCP Generative AI Leader Practice Question 75: Vision API Agent Tool
Google Cloud Generative AI Leader · Test 8 · question 5 of 10
An insurance company is building an agent that lets adjusters upload photos of vehicle damage and receive an automatic list of damage types and their approximate locations in the photo. The company wants to call a pre-built Google Cloud API as a tool for this single-image analysis step, rather than train a custom computer vision model.
Which pre-built API should the agent call for this tool?
Show answer and explanation
Answer:
Cloud Vision API is a pre-built API for analyzing still images: it can detect objects, labels, text, and faces, which fits an agent tool that needs to classify and locate damage in a single uploaded photo. Calling this API as a tool means the company gets image understanding without training or maintaining its own model.
A is wrong because Document AI is purpose-built for extracting structured data from documents like invoices and forms, not general object detection in photos. B is wrong because Video Intelligence API analyzes motion and content across video frames, the wrong tool for a single still photo; Google has also deprecated it as of September 2026, with a shutdown planned for September 2027 and customers directed to migrate to Gemini's multimodal capabilities. D is wrong because Natural Language API works on text input, not images, so it cannot analyze a photo at all.
Reference: Vision AI and Video Intelligence API deprecations