Generative AI Leader Practice Test 25 — 15 Questions

Question 1

A bank is deploying a retrieval-augmented generation application on Vertex AI so that a Gemini model answers employee policy questions using the bank's internal document repository. The team wants the model's responses to cite source documents and to reduce fabricated content. Which two capabilities should they implement to ground the model in the bank's own content? (Choose two.)

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Vertex AI embeddings with vector search over the document corpus. Grounding responses in the bank's own content requires retrieving relevant internal passages and supplying them to the model. Indexing documents with Vertex AI Search or generating embeddings and retrieving them through vector search both provide that context, which improves factual accuracy and supports citations to source documents. Training or safety-filter adjustments do not retrieve proprietary content at inference time and therefore cannot satisfy the citation requirement.

Answer

Gemini safety filters configured to block sensitive topics

Answer

Grounding with Google Search

Answer

Vertex AI Pipelines scheduled retraining of the Gemini model

Question 2

A company's generative AI model is producing biased outputs. What is the most effective mitigation strategy?

Accepted Answer

Fine-tune the model using a balanced, representative dataset and implement output filtering. Fine-tuning on a balanced, representative dataset directly addresses the root cause of biased outputs by correcting the model's learned associations, while output filtering provides a safety net to catch residual bias. This combination is more effective than superficial fixes because it modifies the model's internal weights rather than just masking outputs.

Answer

Use a larger model with more parameters to improve overall accuracy

Answer

Use prompt engineering to instruct the model to avoid biased language

Answer

Increase the diversity of input samples by random sampling

Question 3

A company wants to integrate GenAI into their existing customer relationship management (CRM) system. The CRM is a third-party SaaS application. Which implementation pattern is MOST suitable?

Accepted Answer

API-first integration by calling Vertex AI API from the CRM's custom code. API-first integration allows the CRM to call Google Cloud GenAI APIs without modifying the CRM's core. Workspace add-ons are for Google Workspace, not SaaS CRM.

Answer

Building a standalone AI application and exporting data manually

Answer

Embedding GenAI using Google Workspace add-ons

Answer

Using Apps Script to extend Google Sheets connected to the CRM

Question 4

Refer to the exhibit. What access does the IAM policy grant to developer@example.com?

Accepted Answer

Ability to use Vertex AI models for prediction and view metadata.. The IAM policy grants the 'Vertex AI User' role to developer@example.com, which includes permissions for using models for prediction (e.g., `aiplatform.predict`) and viewing metadata (e.g., `aiplatform.models.list`). This role does not include permissions for deploying or managing models, nor full control over all Vertex AI resources, making option A correct.

Answer

No effective permissions because the role is incorrect.

Answer

Ability to deploy and manage models.

Answer

Full control over all Vertex AI resources.

Question 5

A data analyst wants to use Gemini for Google Workspace to assist with Sheets. Which two capabilities does Duet AI in Sheets provide? (Choose TWO)

Accepted Answer

Generate formula suggestions based on natural language descriptions. Duet AI in Sheets can interpret natural language descriptions of desired calculations and automatically generate the corresponding formula. This leverages Gemini's language understanding to translate user intent into spreadsheet functions, significantly reducing the need to manually recall or construct complex formulas.

Answer

Apply conditional formatting rules based on patterns

Answer

Create charts from images pasted into the sheet

Answer

Write custom Apps Script code automatically

Question 6

A retail company uses the Vertex AI Gemini API to generate product descriptions. Recently, the model started producing factually incorrect statements about product specifications, such as wrong dimensions and materials. Which strategy should be implemented to improve factual accuracy?

Accepted Answer

Use grounding with Vertex AI Search to retrieve verified product data. Grounding with Vertex AI Search connects the Gemini API to a verified, structured data source (e.g., product catalog), enabling the model to retrieve and cite factual specifications rather than relying solely on its training data. This directly addresses the hallucination of wrong dimensions and materials by providing a retrieval-augmented generation (RAG) mechanism that overrides the model's parametric knowledge with authoritative, up-to-date information.

Answer

Enable model versioning to automatically roll back to a previous version

Answer

Fine-tune the model on a dataset of product images and descriptions

Answer

Increase the temperature parameter to 0.9

Question 7

A company is planning an iterative rollout of a GenAI assistant for sales teams. They want to maximize adoption and minimize disruption. Which TWO change management practices should they prioritize? (Choose two.)

Accepted Answer

Provide hands-on training sessions and documentation. Hands-on training and documentation directly address the learning curve and anxiety associated with adopting a GenAI assistant, ensuring users understand how to interact with the model effectively (e.g., prompt engineering basics, interpreting outputs) and reducing friction. This practice aligns with the iterative rollout strategy by providing staged support, which maximizes adoption through competence and confidence rather than coercion.

Answer

Disable the legacy system immediately after GenAI launch

Answer

Roll out to all employees simultaneously to create urgency

Answer

Set a mandatory usage target for each sales rep

Question 8

A company wants to adopt GenAI for contract analysis. They are evaluating build vs. buy. Which TWO factors are MOST important when deciding to build a custom fine-tuned model instead of using a pre-built API? (Choose 2)

Accepted Answer

Desire to keep sensitive contract data within the company's VPC. Need for domain-specific terminology and data privacy requirements are key reasons to build. Cost and speed of deployment favor buy. Model accuracy can be high for both.

Answer

Need to deploy the solution within a week

Answer

Low budget for AI development

Answer

Inability to evaluate model accuracy

Question 9

A company is using Vertex AI to generate report summaries. They want to measure ROI. Which three metrics should they track? (Choose THREE)

Accepted Answer

Error rate (e.g., factual errors in summaries). Error rate directly measures the quality and reliability of generated summaries, which is a key factor in determining ROI. If summaries contain factual errors, they require human correction, reducing the time savings and potentially introducing business risks. Tracking error rate helps quantify the cost of inaccuracies against the benefits of automation.

Answer

Number of employees trained on the system

Answer

Number of active users per month

Question 10

A developer runs this command: `gcloud ai models upload --region=us-central1 --display-name=my-model --artifact-uri=gs://my-bucket/model.pkl`. What is the primary purpose?

Accepted Answer

Upload a model artifact to Model Registry. The command `gcloud ai models upload` uploads a local model artifact (model.pkl stored in Cloud Storage) to the Vertex AI Model Registry. This is used for versioning and managing trained models, not for initiating training, deployment, or pipeline creation. The Model Registry stores the model artifact for later use in deployments or predictions.

Answer

Create a training pipeline

Answer

Deploy a model to an endpoint

Answer

Train a model

Question 11

A legal firm uses a generative AI to draft contracts. They want the output to follow a specific clause structure. Which technique should they use in the prompt?

Accepted Answer

Include a system instruction that defines the required format.. A system instruction (or system message) sets the overall behavior and output format for the generative AI model, effectively constraining it to follow a specific clause structure. This is the most direct and reliable technique for enforcing a predefined format in the prompt, as it operates at the model's instruction-following layer.

Answer

Increase temperature to encourage variance.

Answer

Use grounding to pull from a database of contracts.

Answer

Set stop sequences to end generation at certain points.

Question 12

A retail company wants to build a chatbot that answers product questions and provides personalized recommendations. They have a small labeled dataset and limited ML expertise. Which approach should they take?

Accepted Answer

Use Vertex AI Agent Builder with a pre-built agent and integrate their product catalog via Search and Conversation.. Vertex AI Agent Builder provides a pre-built agent framework that integrates with Search and Conversation, allowing the company to quickly deploy a chatbot using their product catalog without needing extensive ML expertise. This approach leverages Google's foundation models and retrieval-augmented generation (RAG) to answer product questions and generate personalized recommendations, making it ideal for a small labeled dataset and limited ML resources.

Answer

Fine-tune Gemini with their product data using Vertex AI Generative AI Studio.

Answer

Build a custom transformer model using TensorFlow on Vertex AI Workbench.

Answer

Use BigQuery ML to train a classification model on customer queries.

Question 13

A financial services firm needs to fine-tune a large language model on proprietary financial data. They require data to never leave their VPC and need full audit logging. Which Gemini access method should they use?

Accepted Answer

Vertex AI. Vertex AI is the correct access method because it is the only option that allows fine-tuning of Gemini models within a customer's VPC (Virtual Private Cloud) with full audit logging via Cloud Audit Logs. This ensures proprietary financial data never leaves the secure network boundary, meeting strict compliance and data residency requirements.

Answer

Google AI Studio

Answer

Gemini API directly via Cloud Endpoints

Answer

Model Garden in Colab

Question 14

A developer receives the above JSON response from a Vertex AI language model. The output content is correct, but the developer expected the model to not answer geography questions. What should the developer do to prevent the model from responding to geography queries?

Accepted Answer

Add a system instruction to not answer geography questions. Vertex AI does not have a predefined safety filter category for 'Geography'. To prevent the model from answering geography questions, the developer should use a system instruction that explicitly tells the model not to respond to geography queries. System instructions are the appropriate mechanism for custom content restrictions, while safety filters are limited to predefined categories such as toxic, harassment, etc.

Answer

Adjust the safety filter thresholds for the 'Toxic' category

Answer

Enable Vertex AI Grounding with a geography knowledge base

Answer

Configure a safety filter for the 'Geography' category

Question 15

A team is deploying a text generation model for legal document review. They observe that the model occasionally generates factually incorrect legal citations. Which approach best reduces this issue?

Accepted Answer

Implement retrieval-augmented generation (RAG) with a verified legal database.. Retrieval-augmented generation (RAG) with a verified legal database grounds the model in factual, up-to-date sources, directly addressing incorrect citations. Option B (lowering temperature) reduces randomness but does not prevent hallucination. Option C (using a larger model) may not guarantee correctness without proper grounding. Option D (increasing max tokens) has no effect on factual accuracy.

Answer

Lower the temperature to 0.0.

Answer

Use a larger base model.

Answer

Increase the max output tokens.