AI Associate Practice Test 18 — 15 Questions

Question 1

A company wants to ensure their AI model complies with ethical guidelines. Which TWO actions are essential? (Choose two.)

Accepted Answer

Provide human oversight. Option B is correct because human oversight is essential for ethical AI compliance: keeping a human in the loop allows review, intervention, and accountability for model outputs, especially in high-risk or ambiguous cases where automated judgments could cause harm. Option E is correct because documenting model decisions provides the audit trail, traceability, and explainability needed to demonstrate that the AI adheres to ethical guidelines and to support regulatory or internal reviews. The remaining options do not belong: avoiding transparency (A) directly undermines accountability and explainability, using the most complex model (C) does not improve ethics and can reduce interpretability, and automating all decisions (D) removes the human oversight that ethical compliance requires.

Answer

Avoid transparency

Answer

Use the most complex model

Answer

Automate all decisions

Question 2

Refer to the exhibit. An admin built a prediction model for case closure within 24 hours. The model accuracy is 72% with 500 training records. Which change would most likely improve accuracy?

Accepted Answer

Increase the training sample size to 5000 records. Increasing the training sample size from 500 to 5000 records provides the model with more data to learn patterns from, which reduces overfitting and improves generalization. In CRM AI models, larger datasets typically lead to higher accuracy because the algorithm can better capture underlying relationships without being skewed by noise in a small sample.

Answer

Change the outcome to 'Escalated'

Answer

Remove the 'Subject' field from the model

Answer

Add more fields like 'Comments'

Question 3

A company is deploying an AI system to recommend products to customers. To comply with GDPR's right to explanation, which TWO practices should they implement? (Choose 2)

Accepted Answer

Provide a list of the most influential factors for each recommendation. Option B is correct because GDPR's right to explanation (Articles 13-15 and Recital 71) requires that data subjects be given meaningful information about the logic involved in automated decision-making, and listing the most influential factors per recommendation directly satisfies this transparency requirement. Option E is correct because GDPR Article 22 grants individuals the right not to be subject to solely automated decisions with significant effects and provides the right to obtain human intervention, so offering a human review process when a customer requests an explanation operationalizes that safeguard. Option A is not required by the right to explanation; GDPR retention is governed by the storage limitation principle (Article 5(1)(e)), which favors keeping data no longer than necessary, not a fixed 10-year period. Option C addresses consent/opt-out rights under Articles 21-22 but is not the specific mechanism for delivering an explanation. Option D concerns data minimization and anonymization, which reduce privacy risk but do not by themselves fulfill the right to explanation for automated decisions.

Answer

Store all recommendation logs for at least 10 years

Answer

Allow customers to opt out of all AI recommendations

Answer

Use only anonymized data for recommendations

Question 4

A developer is building an autonomous AI agent with Agentforce. They need the agent to perform actions in Salesforce, such as updating records and sending emails. How should they define these capabilities in Agent Builder?

Accepted Answer

Create actions in Agent Builder, specifying the operation and parameters. In Agent Builder, actions are the mechanism that defines what an autonomous AI agent can do in Salesforce, such as updating records or sending emails. Each action specifies the operation (e.g., a standard or custom action) and its parameters, allowing the agent to execute precise tasks without additional coding.

Answer

Define topics that correspond to each action

Answer

Use Prompt Builder to create prompts for each action

Answer

Write Apex triggers to handle agent requests

Question 5

A Salesforce admin implements Einstein Bots for customer service. To ensure the bot does not use biased language, what should the admin do?

Accepted Answer

Review training data for representativeness. Bias in an Einstein Bot originates from the data used to train its language understanding models, not from the responses it returns at runtime. Reviewing training data for representativeness ensures the utterances, intents, and entities reflect a diverse and balanced set of user inputs, which directly reduces biased intent classification and response selection. This is the only option that addresses the root cause of biased language in the model.

Answer

Use only predefined responses

Answer

Disable the bot for sensitive topics

Answer

Monitor conversations regularly

Question 6

A sales team is using Einstein Lead Scoring and notices that leads from a certain geographic region are consistently scored lower, even when the lead's profile matches high-performing customers from other regions. What is the most likely cause and recommended first step?

Accepted Answer

The model may be biased due to underrepresented data from that region in the training set. Audit the model for bias and review training data demographics.. When a model systematically scores one geographic segment lower despite comparable profiles, the most likely cause is underrepresentation or bias in the training data for that region. The correct first step is to audit the model for bias and review the training data demographics, because Einstein Lead Scoring learns patterns from historical data and will reproduce any historical skew. Auditing before retraining ensures the fix addresses the root cause rather than amplifying it.

Answer

The model is accurate; the region is genuinely underperforming. The team should accept the scores.

Answer

The region has lower quality leads, so the scores are correct. The team should focus on other regions.

Answer

Retrain the model with more data from that region, even if it means duplicating records.

Question 7

Which Salesforce Einstein feature provides automated statistical analysis of data, generates stories in natural language, and offers improvement suggestions in a waterfall chart?

Accepted Answer

Einstein Discovery. Einstein Discovery is the correct answer because it is the Salesforce AI feature specifically designed to perform automated statistical analysis on data, generate natural language narratives (stories) that explain key insights, and provide actionable improvement suggestions visualized in a waterfall chart. Unlike other Einstein features, Discovery focuses on surfacing hidden patterns and recommending specific actions to improve business outcomes.

Answer

Einstein GPT

Answer

Einstein Analytics

Answer

Einstein Prediction Builder

Question 8

Refer to the exhibit. The prediction API returns a probability of 0.85 for the label 'High Value'. What does this value represent?

Accepted Answer

The confidence score that this lead is 'High Value'. The prediction API returns a probability of 0.85 for the label 'High Value'. In machine learning classification, this output represents the model's confidence score—the estimated probability that the input instance belongs to the specified class. It is not a direct measure of conversion likelihood, model accuracy, or F1 score; it is the raw posterior probability assigned by the model to the 'High Value' label.

Answer

The likelihood that this lead will convert

Answer

The F1 score of the model for this prediction

Answer

The model's accuracy on the training set

Question 9

A developer is creating a custom AI model on Salesforce. To ensure the model is fair across demographic groups, which activity should be included in the development process?

Accepted Answer

Bias testing using a diverse test dataset.. Bias testing using diverse datasets directly evaluates fairness across groups.

Answer

Feature selection using correlation matrix.

Answer

Cross-validation to avoid overfitting.

Answer

Hyperparameter tuning with grid search.

Question 10

A developer is building an Einstein Bot that needs to understand when a customer says 'I want to return a purchase' and route them to the returns process. How should they configure the bot?

Accepted Answer

Create an intent called 'Return Purchase' and train it with sample phrases. In Einstein Bots, intents represent the customer's goal, and entities capture specifics. 'Return purchase' is an intent, not an entity. The bot uses NLP to match utterances to intents.

Answer

Create a dialogue that triggers on the exact phrase 'I want to return a purchase'

Answer

Create an entity called 'Return Purchase' and map it to a dialogue

Answer

Use an intent called 'Customer Service' and a custom entity for return

Question 11

A service agent needs to quickly find a relevant knowledge article while working on a case. Which Einstein feature can automatically suggest articles based on the case details?

Accepted Answer

Einstein Article Recommendations. Einstein Article Recommendations is the correct answer because it is the specific Einstein feature designed to automatically surface relevant knowledge articles based on the context of a case, such as subject, description, and product. It uses natural language processing (NLP) to match case details against article content, providing agents with immediate, relevant suggestions without manual search.

Answer

Einstein Prediction Builder

Answer

Einstein Case Classification

Answer

Einstein Next Best Action

Question 12

An AI system used for medical diagnosis has been shown to have lower accuracy for certain ethnic groups. The development team is considering releasing it anyway because most patients are from the majority group. Which ethical principle is most compromised?

Accepted Answer

Fairness. The scenario describes an AI system that performs worse for certain ethnic groups, yet the team plans to release it anyway because the majority group is unaffected. This directly violates the principle of fairness, which requires that AI systems do not discriminate or perpetuate bias against any group. Releasing a model with known accuracy disparities without mitigation prioritizes overall performance over equitable treatment, compromising fairness.

Answer

Transparency

Answer

Accountability

Answer

Privacy

Question 13

Which TWO actions are recommended when preparing data for an Einstein Prediction Builder model?

Accepted Answer

Ensure the data set contains at least 500 records with the outcome field populated.. Einstein Prediction Builder requires a minimum of 500 records with the outcome field populated to ensure statistical significance and reliable model training. Fewer records can lead to overfitting or insufficient pattern recognition, making the model less accurate.

Answer

Include all available fields on the object, even if unrelated.

Answer

Use external data sources and upload CSV files without any preprocessing.

Answer

Include data from the last two days only for the most current trends.

Question 14

A financial services firm deployed an AI model to automate loan approvals. The model was trained on historical loan data from the past 10 years, which shows that applicants from certain zip codes have higher default rates. After six months, the company's compliance team receives complaints that applicants from predominantly low-income neighborhoods are being rejected at a much higher rate than applicants from affluent areas, even when their financial profiles are similar. The model's overall accuracy remains high (95%), and the loan default rate has decreased by 15% since deployment. The company wants to address the ethical concerns without sacrificing performance. Which course of action should the company take?

Accepted Answer

Retrain the model with a balanced dataset that includes more examples from underrepresented neighborhoods and enforce fairness constraints.. The model is exhibiting disparate impact via proxy variables (zip code correlates with socioeconomic status), so the fix is to retrain with a more balanced dataset and enforce fairness constraints, which addresses the root cause while preserving performance. Simply removing zip code doesn't eliminate correlated proxies, and threshold manipulation is itself discriminatory. A balanced retraining approach targets both fairness and accuracy.

Answer

Remove the zip code feature from the model inputs.

Answer

Adjust the approval threshold lower only for applicants from low-income neighborhoods.

Answer

Continue using the existing model since it has high accuracy and reduces defaults.

Question 15

What is the term for when an AI model produces confident but incorrect information, often in generative AI?

Accepted Answer

Hallucination. Hallucination refers to the phenomenon where a generative AI model produces output that is fluent and confident but factually incorrect or fabricated. This is distinct from bias (systematic skew in outputs), overfitting (model memorizes training data), and underfitting (model fails to capture patterns). Option C is correct.

Answer

Bias

Answer

Overfitting

Answer

Underfitting