Courseiva

Salesforce AI Associate AI Associate (AI Associate) — Questions 676–750

1000 questions total · 14pages · All types, answers revealed

Take a mock examExam hub
Page 9

Page 10 of 14

Page 11
676
MCQmedium

A data scientist is building a model for credit scoring. They have access to a dataset with historical bias. What should they do?

A.Apply fairness constraints during training.
B.Discard all biased variables.
C.Use the data as is because it reflects reality.
D.Use a more complex model to reduce bias.
AnswerA

Helps mitigate bias.

Why this answer

Option B is correct because applying fairness constraints during training helps mitigate bias. Option A is wrong using biased data as is perpetuates bias. Option C is wrong discarding all biased variables may remove useful information and doesn't guarantee fairness.

Option D is wrong increasing model complexity can amplify bias.

Full explanation →
677
Multi-Selecteasy

A marketing team uses Einstein Send Time Optimization to determine the best time to send emails. To ensure the company follows Salesforce's Trusted AI principle of Honesty, which THREE practices should be adopted?

Select 3 answers
A.Automate all email sends without human intervention to maintain consistency
B.Tell recipients that the email timing was chosen by an AI algorithm
C.Provide a brief explanation of the factors that influenced the recommendation
D.Document the model's limitations (e.g., it may not account for time zones)
E.Claim the model is 100% accurate in predicting optimal send times
AnswersB, C, D

Transparency about AI use is part of honesty.

Why this answer

Honesty means being transparent about AI limitations and capabilities. Providing documentation, clearly communicating that predictions are AI-based, and explaining the influencing factors align with honesty. Automating without review and overpromising accuracy are contrary.

Full explanation →
678
MCQeasy

Which Einstein feature records call recordings and provides analysis on keywords, talk-time metrics, and next steps?

A.Einstein Email Insights
B.Einstein Conversation Insights
C.Einstein Activity Capture
D.Einstein Discovery
AnswerB

Conversation Insights provides call recording analysis and metrics.

Why this answer

Einstein Conversation Insights is the correct feature because it is specifically designed to record and analyze sales calls, providing transcriptions, keyword spotting, talk-time metrics, and automated next steps. It uses natural language processing (NLP) to surface insights from conversational data, directly matching the question's requirements.

Exam trap

The trap here is that candidates confuse Einstein Conversation Insights with Einstein Email Insights, assuming both handle communication analysis, but only Conversation Insights processes real-time call recordings and audio metrics.

How to eliminate wrong answers

Option A is wrong because Einstein Email Insights analyzes email interactions, not call recordings, and focuses on email engagement metrics like open rates and reply patterns. Option C is wrong because Einstein Activity Capture syncs calendar events and emails from Exchange or Gmail into Salesforce, but does not record or analyze call audio. Option D is wrong because Einstein Discovery is a predictive analytics and AI-powered recommendation engine for data patterns, not a tool for recording or analyzing call conversations.

Full explanation →
679
MCQhard

Refer to the exhibit. A developer configured a grounding policy for Einstein GPT. What is the effect of the fallbackBehavior set to 'USE_MODEL_KNOWLEDGE'?

A.The AI will rely on its pre-trained knowledge when no grounding data meets the relevance threshold
B.The AI will return all retrieved grounding data in the response
C.The AI will not generate a response if no grounding data is found
D.The AI will ignore the grounding policy and only use model knowledge
AnswerA

This is the defined fallback: use the model's internal knowledge if grounding data is insufficient.

Why this answer

When fallbackBehavior is set to 'USE_MODEL_KNOWLEDGE', the Einstein GPT grounding policy instructs the AI to fall back to its pre-trained (model) knowledge if the retrieved grounding data does not meet the configured relevance threshold. This ensures the AI still generates a response based on its internal training rather than returning no answer or ignoring the policy entirely.

Exam trap

Salesforce often tests the distinction between 'fallback' and 'ignore' — the trap here is assuming 'USE_MODEL_KNOWLEDGE' means the grounding policy is disregarded, when in fact it is a controlled fallback within the policy's logic.

How to eliminate wrong answers

Option B is wrong because setting fallbackBehavior to 'USE_MODEL_KNOWLEDGE' does not cause the AI to return all retrieved grounding data; it only triggers a fallback to model knowledge when relevance is insufficient. Option C is wrong because the AI will still generate a response using its pre-trained knowledge rather than refusing to respond when no grounding data meets the threshold. Option D is wrong because the grounding policy is not ignored; the fallback is a defined behavior within the policy, not a bypass of the policy itself.

Full explanation →
680
MCQmedium

A lead scoring model trained on historical sales data is found to assign lower scores to leads from certain postal codes. What is the MOST likely cause?

A.The model's algorithm is inherently biased against certain regions
B.The training data contains biased outcomes from past human decisions
C.The model overfits to the training data
D.The model was not trained long enough
AnswerB

Historical bias in the data leads the model to replicate those biases.

Why this answer

The model assigns lower scores to leads from certain postal codes because the training data reflects historical human biases, such as sales representatives prioritizing leads from affluent areas. Machine learning models learn patterns from the data they are trained on; if past sales decisions were biased against certain regions, the model will replicate those biases. This is a classic case of bias in training data leading to biased model outcomes, not an inherent flaw in the algorithm itself.

Exam trap

Cisco often tests the misconception that algorithmic bias is caused by the model's internal logic or training duration, whereas the root cause is almost always biased training data reflecting past human decisions.

How to eliminate wrong answers

Option A is wrong because algorithms are not inherently biased; bias arises from the data or feature engineering, not from the algorithm's design. Option C is wrong because overfitting would cause the model to perform poorly on new data by memorizing noise, not systematically assign lower scores to specific postal codes. Option D is wrong because insufficient training would result in underfitting, where the model fails to capture patterns, not produce consistent region-based score disparities.

Full explanation →
681
MCQhard

You are a Salesforce AI Specialist at a mid-sized manufacturing company. The company uses Einstein Lead Scoring to prioritize leads. The model was trained on historical lead data and has been in production for three months. Recently, the sales team reports that high-scoring leads are not converting as expected. You investigate and find that the model's data source includes leads from the past 18 months. However, six months ago, the company changed its lead qualification process: they started requiring a demo before scoring leads as 'qualified.' As a result, the definition of a converted lead changed. What is the best course of action to improve model performance?

A.Manually adjust the model's prediction threshold to account for the new process
B.Retrain the model using only leads from the last six months after the process change
C.Remove the 'Demo Scheduled' field from the model to avoid bias
D.Add more historical leads from before the process change to increase data volume
AnswerB

This ensures the model learns from data that reflects the current conversion criteria.

Why this answer

Option B is correct because the change in lead qualification process six months ago introduced a data distribution shift (concept drift), making older leads no longer representative of the current conversion behavior. Retraining the model on only the last six months of data aligns the training set with the new definition of a 'converted lead,' allowing Einstein Lead Scoring to learn the updated patterns and improve prediction accuracy.

Exam trap

The trap here is that candidates may think adjusting the threshold (Option A) is sufficient, but they fail to recognize that a change in the definition of the target variable requires retraining on a representative dataset, not just tuning a post-processing parameter.

How to eliminate wrong answers

Option A is wrong because manually adjusting the prediction threshold does not address the underlying change in the definition of a converted lead; it only shifts the cutoff for scoring, which cannot compensate for a fundamentally different target variable. Option C is wrong because removing the 'Demo Scheduled' field does not solve the problem—the issue is the change in the conversion definition, not bias from that field; in fact, the field may now be more predictive under the new process. Option D is wrong because adding more historical leads from before the process change would exacerbate the data mismatch, as those leads follow the old qualification rules and would dilute the model's ability to learn the current conversion patterns.

Full explanation →
682
MCQhard

An AI system for medical diagnosis is trained on data from one region. When deployed globally, it performs poorly. This is an issue of?

A.Accountability
B.Overfitting
C.Privacy violation
D.Generalizability
AnswerD

Lack of generalizability means the model fails on data from different distributions.

Why this answer

Option B is correct because poor generalization across regions indicates the model is not generalizable. Option A is wrong overfitting would cause poor performance on new data within the same region. Option C is wrong privacy relates to data protection, not performance.

Option D is wrong accountability is about responsibility, not technical limitation.

Full explanation →
683
MCQmedium

A data scientist needs to feed customer interaction data into Einstein Discovery for predictive analysis. Which data format is required?

A.CSV
B.XML
C.Parquet
D.JSON
AnswerA

CSV is the required format for Einstein Discovery.

Why this answer

Option C is correct because Einstein Discovery typically requires data in CSV format. Option A is wrong because JSON is not the standard input format for Einstein Discovery. Option B is wrong because XML is not commonly used.

Option D is wrong because Parquet is a columnar storage format not directly supported.

Full explanation →
684
Multi-Selecteasy

Which TWO of the following are required for GDPR compliance when using AI with personal data?

Select 2 answers
A.Storing data indefinitely for future AI training
B.Obtaining explicit consent from users before processing their data
C.Selling user data to third parties for AI model improvement
D.Providing users the ability to request deletion of their data
E.Processing data without informing users
AnswersB, D

Consent is a lawful basis for processing under GDPR.

Why this answer

Options A and C are correct: Obtaining explicit consent and enabling data deletion (right to erasure) are GDPR requirements. Option B (Storing data indefinitely) violates storage limitation. Option D (Selling data without permission) is illegal.

Option E (Processing data without consent) is not allowed.

Full explanation →
685
Multi-Selecthard

Which THREE are key dimensions of data quality that directly impact AI model performance?

Select 3 answers
A.Consistency.
B.Timeliness.
C.Accuracy.
D.Data volume.
E.Completeness.
AnswersA, C, E

Inconsistent data (e.g., different formats) confuses models and degrades accuracy.

Why this answer

Consistency is a key dimension of data quality because AI models rely on stable patterns in the data. If the same entity is represented differently across records (e.g., 'NY' vs 'New York'), the model may learn incorrect correlations, leading to degraded prediction accuracy and unreliable outputs.

Exam trap

Salesforce often tests the distinction between data quality dimensions and data quantity metrics, so candidates mistakenly select 'data volume' thinking more data always improves AI performance, when in fact the exam focuses on accuracy, completeness, and consistency as the three critical quality dimensions.

Full explanation →
686
MCQhard

A support center wants to use Einstein Case Classification to automatically assign categories to incoming cases. They have historical case data with the 'Type' field populated for 70% of cases, 'Priority' for 50%, and 'Reason' for 30%. They want to classify on 'Type' and 'Reason'. What is the best approach to maximize model accuracy?

A.Use Einstein Prediction Builder instead, which can handle multi-class classification for both fields.
B.Build a single model that predicts both Type and Reason simultaneously to leverage all available data.
C.Start by training a model for Type only, since it has more populated records, then train a model for Reason once more data is accumulated.
D.Create two separate models: one for Type using all records with Type populated, and one for Reason using all records with Reason populated.
AnswerC

Starting with Type gives a larger training set (70% of records), likely meeting the 1500 record minimum. Reason can be added later when more records have that field populated.

Why this answer

Einstein Case Classification requires a minimum of 1500 records with the target field populated. Prioritizing the field with more populated records (Type) ensures a larger training set, improving accuracy. Reason can be added later once more data is available.

Full explanation →
687
MCQeasy

A company wants to use Einstein Prediction Builder to predict which customers are likely to churn. They have a dataset that includes customers' names, email addresses, and detailed purchase history. According to the data minimisation principle, which fields should be included in the model?

A.Only the purchase history
B.Only the email addresses, since they are unique identifiers
C.All three fields because more data improves accuracy
D.Names and purchase history, but not email addresses
AnswerA

Purchase history is relevant; names and emails are unnecessary PII. Excluding them minimizes data use and privacy risk.

Why this answer

Data minimisation means using only the data necessary for the prediction. Purchase history is relevant to churn prediction, while names and email addresses are personally identifiable information (PII) that are not needed for the model and should be excluded to minimize privacy risks.

Full explanation →
688
MCQeasy

A nonprofit organization wants to automatically recommend relevant knowledge articles to service agents while they are working on a case. Which feature should they enable?

A.Einstein Next Best Action
B.Einstein Recommendation Builder
C.Einstein Article Recommendations
D.Einstein Case Classification
AnswerC

This feature recommends articles to agents during case handling.

Why this answer

Einstein Article Recommendations is the correct feature because it specifically uses AI to analyze the context of a service case (such as subject, description, and product) and automatically surfaces relevant knowledge articles to the agent without requiring manual search. This directly matches the requirement of recommending articles while the agent works on a case.

Exam trap

The trap here is that candidates often confuse Einstein Next Best Action (which can also surface recommendations) with Einstein Article Recommendations, but Next Best Action is a broader tool for any type of recommendation (e.g., offers, next steps) and is not specifically optimized for knowledge articles in a service case context.

How to eliminate wrong answers

Option A is wrong because Einstein Next Best Action is designed to deliver personalized recommendations, prompts, or offers to customers or agents based on rules and AI, but it is not specifically built for recommending knowledge articles within a service case context. Option B is wrong because Einstein Recommendation Builder is a tool for creating custom AI-powered recommendations for products, content, or actions, but it requires manual configuration and is not an out-of-the-box feature for automatic article recommendations in service cases. Option D is wrong because Einstein Case Classification uses AI to automatically classify cases (e.g., by type, priority, or route) based on historical data, but it does not recommend knowledge articles to agents.

Full explanation →
689
MCQmedium

A company wants to use Einstein to automatically log emails and events from their email system into Salesforce without manual user action. Which feature should be enabled?

A.Einstein Email Insights
B.Einstein Conversation Insights
C.Einstein Discovery
D.Einstein Activity Capture
AnswerD

Activity Capture automatically logs emails and events from connected email accounts.

Why this answer

Einstein Activity Capture automatically logs emails and events to Salesforce based on sync settings.

Full explanation →
690
MCQhard

A company is concerned about the data minimization principle when using AI to predict customer lifetime value. Which approach aligns with this principle?

A.Include all available fields to maximize model accuracy
B.Use only fields directly relevant to purchase history and engagement, excluding PII like social security numbers
C.Obtain explicit consent for every field used
D.Anonymize all data before training, including purchase amounts
AnswerB

Using only relevant fields reduces privacy risk and aligns with minimization.

Why this answer

Option B aligns with the data minimization principle by restricting data collection to only fields directly relevant to the prediction task (purchase history and engagement) and explicitly excluding personally identifiable information (PII) like social security numbers. This reduces privacy risk and complies with regulations such as GDPR and CCPA, which require that data collected be adequate, relevant, and limited to what is necessary for the processing purpose.

Exam trap

Cisco often tests the misconception that obtaining consent or anonymizing data automatically satisfies data minimization, when in fact these measures address different principles (consent and data security) and do not reduce the scope of data collected.

How to eliminate wrong answers

Option A is wrong because including all available fields violates data minimization by collecting irrelevant or excessive data, which increases privacy risk and may introduce bias or overfitting without improving model accuracy. Option C is wrong because obtaining explicit consent for every field used addresses consent requirements but does not inherently satisfy data minimization; consent does not justify collecting unnecessary data. Option D is wrong because anonymizing all data before training, including purchase amounts, may still violate data minimization if the anonymized data includes fields that are not necessary for the prediction task, and anonymization does not reduce the volume of data collected.

Full explanation →
691
MCQmedium

A service agent is working on a case and needs to quickly find relevant knowledge articles without searching manually. Which Einstein feature can automatically suggest articles based on the case details?

A.Einstein Next Best Action
B.Einstein Search
C.Einstein Article Recommendations
D.Einstein Case Classification
AnswerC

Article Recommendations suggests relevant knowledge articles automatically.

Why this answer

Einstein Article Recommendations is the correct feature because it automatically suggests relevant knowledge articles based on the case details, such as subject, description, and product. This eliminates the need for manual search by leveraging AI to match case context with article content.

Exam trap

The trap here is that candidates may confuse Einstein Article Recommendations with Einstein Case Classification, because both use case details, but one suggests articles while the other assigns field values.

How to eliminate wrong answers

Option A is wrong because Einstein Next Best Action recommends the next best action for an agent to take (e.g., a guided process or offer), not knowledge articles. Option B is wrong because Einstein Search is a natural language search tool that requires the user to input a query, whereas the question specifies automatic suggestions without manual searching. Option D is wrong because Einstein Case Classification automatically assigns case fields like type or priority based on case details, but it does not suggest knowledge articles.

Full explanation →
692
MCQmedium

A company uses Einstein Lead Scoring and finds that leads with high scores are not converting. What should the admin do to improve prediction accuracy?

A.Increase the scoring model's maximum score
B.Retrain the model with more recent conversion data
C.Disable field-level security for scoring fields
D.Lower the lead conversion threshold
AnswerB

Retraining with current data improves the model's relevance.

Why this answer

Option B is correct because retraining the model with more recent conversion data allows the Einstein Lead Scoring model to adapt to changing patterns in lead behavior and conversion criteria. When high-scoring leads fail to convert, it indicates that the historical data used to train the model no longer reflects current conversion dynamics, so refreshing the training dataset improves prediction accuracy by aligning the model with recent outcomes.

Exam trap

Salesforce often tests the misconception that adjusting score thresholds or model parameters can fix accuracy issues, when the real solution is to refresh the training data to reflect current conversion patterns.

How to eliminate wrong answers

Option A is wrong because increasing the maximum score does not address the underlying data mismatch; it merely scales the output range without improving the model's ability to distinguish converters from non-converters. Option C is wrong because disabling field-level security for scoring fields would expose sensitive data and violate Salesforce security best practices, and it does not affect the model's training data or prediction logic. Option D is wrong because lowering the lead conversion threshold artificially inflates conversion rates without fixing the model's accuracy; it treats the symptom (low conversion) rather than the cause (stale training data).

Full explanation →
693
MCQhard

Refer to the exhibit. A company configures a Prompt Builder policy for Einstein GPT. What is the primary role of the 'checkPromptOutput' flag?

A.To log all prompts for audit purposes.
B.To scan the generated text against the banned words list.
C.To limit the total number of tokens in the generated response.
D.To send the output to a human reviewer before sending.
AnswerB

checkPromptOutput likely enables content scanning.

Why this answer

The 'checkPromptOutput' flag in a Prompt Builder policy for Einstein GPT is specifically designed to scan the generated text against a banned words list. This ensures that the AI output does not contain prohibited or sensitive terms, aligning with ethical and compliance requirements. It is a content filtering mechanism, not a logging, token-limiting, or human-review function.

Exam trap

The trap here is that candidates often confuse content filtering (banned words scanning) with broader safety mechanisms like logging, token limits, or human review, because all are related to output control but serve distinct purposes.

How to eliminate wrong answers

Option A is wrong because logging prompts for audit purposes is typically handled by separate audit trail or logging configurations, not the 'checkPromptOutput' flag which focuses on real-time content scanning. Option C is wrong because limiting the total number of tokens in the generated response is controlled by token limit parameters or max tokens settings, not by a flag that checks for banned words. Option D is wrong because sending output to a human reviewer before sending is a human-in-the-loop workflow, often managed by approval policies or review queues, not by the 'checkPromptOutput' flag which automates filtering without human intervention.

Full explanation →
694
MCQhard

A data architect is designing a data model for Einstein Discovery. The data includes categorical variables with high cardinality (e.g., postal codes). What is the best practice to handle such features?

A.Encode them as one-hot vectors.
B.Exclude them from the model.
C.Use the raw values without transformation.
D.Group them into higher-level categories (e.g., region).
AnswerD

Reduces cardinality while preserving signal.

Why this answer

Grouping high-cardinality categories into broader categories reduces overfitting and improves model stability.

Full explanation →
695
Multi-Selectmedium

A Salesforce admin is configuring Einstein Next Best Action. Which TWO actions demonstrate ethical AI practices? (Choose two.)

Select 2 answers
A.Allow the AI to automatically take actions without human approval for all recommendations
B.Use the same recommendation for all customers to ensure fairness
C.Keep the AI's reasoning hidden from business users to avoid confusion
D.Regularly review recommendation logs for patterns of bias or unfair treatment
E.Set up human review for recommendations that involve sensitive decisions about customers
AnswersD, E

Ongoing monitoring detects bias.

Why this answer

Option B (Set up human review for recommendations that involve sensitive decisions) is correct for accountability. Option D (Regularly review recommendation logs for patterns of bias) is correct for monitoring. Option A (Allow the AI to act autonomously without human approval) reduces accountability.

Option C (Use the same recommendation for all customers) ignores personalization ethics. Option E (Hide the AI's reasoning from business users) violates transparency.

Full explanation →
696
MCQmedium

A company has deployed an AI-powered chatbot to handle customer service inquiries. The chatbot is designed to answer frequently asked questions and escalate complex issues to human agents. Which action best aligns with ethical AI principles regarding transparency?

A.Configure the chatbot to answer all queries without human intervention to maximize efficiency.
B.Use historical customer interaction data without auditing for bias to train the chatbot's responses.
C.Program the chatbot to identify itself as a human agent to build trust with customers.
D.Clearly disclose that the chatbot is an AI and provide an option for customers to switch to a human agent.
AnswerD

This ensures transparency by informing customers of the AI nature and respecting user autonomy with an opt-in for human assistance.

Why this answer

Option C is correct because transparency requires disclosing that the interaction is with an AI and allowing customers to opt for human assistance. Option A is wrong because it may not be feasible or ethical to force the chatbot to handle all queries. Option B is wrong because misrepresenting the AI as human is deceptive.

Option D is wrong because ignoring bias in training data is unethical.

Full explanation →
697
MCQmedium

A company is deploying Einstein Article Recommendations on its customer portal. They want to ensure customers know that recommendations are AI-generated. Which action aligns with the Salesforce Trusted AI Principle of Honesty?

A.Provide a chatbot that can explain the recommendations
B.Use a generic label like 'Recommended for you'
C.Include a note stating 'Recommended by Einstein AI' on each recommendation
D.Do not mention AI; just show the recommendations
AnswerC

Clear disclosure that recommendations are AI-generated is honest and transparent, aligning with the principle of Honesty.

Why this answer

Option C is correct because the Salesforce Trusted AI Principle of Honesty requires transparency about AI-generated content. By explicitly stating 'Recommended by Einstein AI' on each recommendation, the company clearly discloses that the recommendations are AI-generated, building trust with customers. This aligns with the principle's focus on avoiding deception and ensuring users understand the nature of the content they are seeing.

Exam trap

Cisco often tests the distinction between indirect transparency (like a chatbot) and direct, upfront disclosure (like a label), leading candidates to choose an option that seems helpful but does not satisfy the specific requirement of the Honesty principle.

How to eliminate wrong answers

Option A is wrong because providing a chatbot that can explain the recommendations does not directly disclose that the recommendations are AI-generated; it only offers an explanation after the fact, which fails to meet the upfront transparency required by the Honesty principle. Option B is wrong because using a generic label like 'Recommended for you' is ambiguous and does not explicitly state that the recommendations are AI-generated, which violates the Honesty principle by potentially misleading customers into thinking the recommendations are human-curated. Option D is wrong because not mentioning AI at all is a direct violation of the Honesty principle, as it conceals the AI-generated nature of the recommendations, undermining customer trust and transparency.

Full explanation →
698
Multi-Selectmedium

An organization wants to use Einstein GPT to generate case summaries and draft knowledge articles for service agents. Which TWO Einstein GPT products should they enable? (Choose two)

Select 2 answers
A.Prompt Builder
B.Service GPT
C.Einstein Copilot
D.Sales GPT
E.Einstein Article Recommendations
AnswersA, B

Prompt Builder is used to create and manage prompt templates that power GPT features, including case summaries and article drafts.

Why this answer

Service GPT includes case summary and knowledge article draft capabilities.

Full explanation →
699
Multi-Selecthard

A retail company uses Einstein Article Recommendations to suggest knowledge articles to customer service agents. To ensure compliance with the Salesforce Trusted AI principle of Safety, which THREE measures should the company implement?

Select 3 answers
A.Log all recommendations without setting up any review process
B.Use grounding to connect the AI recommendations to the company's CRM data
C.Enable toxicity detection in the Einstein Trust Layer to filter inappropriate content
D.Hide the model's confidence score from agents to avoid over-reliance
E.Require human agents to review and approve recommendations before sharing with customers
AnswersB, C, E

Grounding ensures recommendations are based on vetted data, reducing the chance of unsafe outputs.

Why this answer

Safety involves preventing harmful outputs. Human review, toxicity detection, and using grounding to ensure relevance all mitigate harmful recommendations. Hiding the model's confidence score reduces transparency, and logging without review does not prevent harm.

Full explanation →
700
Multi-Selecthard

Which TWO considerations are critical when planning data labeling for a computer vision project in a regulated industry?

Select 2 answers
A.Data storage location for label files
B.Mitigating labeler bias to ensure fairness
C.Compliance with data privacy regulations (e.g., GDPR)
D.Labeling timeline and budget constraints
E.Choosing between bounding boxes and segmentation masks
AnswersB, C

Bias can affect model fairness and regulatory requirements.

Why this answer

Option B is correct because labeler bias can introduce systematic errors into the training data, leading to models that perform unfairly or inaccurately across different demographic groups. In regulated industries, such bias can violate anti-discrimination laws and regulatory standards, making its mitigation a critical planning consideration.

Exam trap

Salesforce often tests the distinction between operational details (like storage location or annotation type) and critical regulatory or ethical considerations, leading candidates to choose technically valid but non-critical options like A or E.

Full explanation →
701
MCQeasy

A sales team uses an AI tool to recommend products to customers. The tool recommends high-commission products over what best fits the customer. Which ethical principle is being violated?

A.Privacy
B.Transparency
C.Fairness
D.Accountability
AnswerC

Recommending based on commission rather than customer need is unfair.

Why this answer

Option B is correct because fairness requires that recommendations benefit the customer, not just the seller. Option A is wrong as transparency is about openness, not directly violated here. Option C is wrong because accountability involves responsibility, which may also be an issue but fairness is primary.

Option D is wrong because privacy is about data protection.

Full explanation →
702
Multi-Selectmedium

A company wants to use Agentforce to create an autonomous AI agent that can handle customer service inquiries. Which TWO components must be configured in Agent Builder?

Select 2 answers
A.Topics
B.Prompt Templates
C.Entities
D.Actions
E.Intents
AnswersA, D

Topics define the subject areas the agent can handle.

Why this answer

A is correct because Topics in Agent Builder define the high-level categories of customer inquiries (e.g., 'Billing', 'Returns') that the autonomous agent can handle. They are mandatory for structuring the agent's conversation flow and routing logic. Without Topics, the agent would have no defined scope of work.

Exam trap

The trap here is that candidates confuse the required components of Agent Builder (Topics and Actions) with similar concepts from other Salesforce AI features, such as Intents from Einstein Bots or Prompt Templates from Einstein Generative AI.

Full explanation →
703
MCQmedium

An admin sets up Einstein Sentiment scoring for case comments. After a week, they notice that most comments are scored as 'Neutral' even when customer sentiment is clearly negative. What should the admin check first?

A.Increase the model confidence threshold to reduce false positives
B.Switch to Einstein Bot for sentiment analysis
C.Verify that case comments contain at least 50 words each
D.Retrain the sentiment model with industry-specific training data
AnswerD

Custom training improves relevance to domain language and expressions.

Why this answer

Option D is correct because Einstein Sentiment is a pre-trained model that may not accurately interpret sentiment in industry-specific contexts. Retraining the model with domain-specific training data (e.g., using Salesforce's Intent and Sentiment API with custom datasets) adjusts the model to recognize sentiment nuances in that industry, improving accuracy beyond the generic baseline.

Exam trap

Salesforce often tests the misconception that adjusting confidence thresholds or switching tools can fix model accuracy issues, when the correct first step is to retrain the model with relevant data.

How to eliminate wrong answers

Option A is wrong because increasing the model confidence threshold would make the model more conservative, likely increasing 'Neutral' scores rather than reducing them; it addresses false positives, not the underlying issue of misclassifying clearly negative sentiment. Option B is wrong because Einstein Bot is a chatbot tool for automating conversations, not a sentiment analysis engine; it cannot replace or fix the sentiment scoring model. Option C is wrong because Einstein Sentiment does not require a minimum word count of 50; it can analyze shorter comments, and the issue is model accuracy, not input length.

Full explanation →
704
MCQhard

An administrator is configuring Einstein Forecasting. They notice the AI forecast differs significantly from the manager's commit. The manager wants to understand why the AI forecast is lower. What should the administrator do?

A.Run an Einstein Discovery story on the forecast data
B.Disable Einstein Forecasting and rely solely on manager commit
C.Check the AI forecast explanation, which shows key drivers like historical win rate and pipeline changes
D.Adjust the AI forecast manually to match the commit
AnswerC

The explanation highlights what drives the AI forecast, helping to reconcile differences.

Why this answer

Option C is correct because Einstein Forecasting provides an AI forecast explanation that details the key drivers influencing the prediction, such as historical win rates and pipeline changes. This explanation allows the administrator to transparently show the manager why the AI forecast is lower, addressing the discrepancy without disabling or manually overriding the AI model.

Exam trap

The trap here is that candidates may think Einstein Discovery is the correct tool for explaining forecasts, but it is designed for broader data exploration, not for providing per-forecast driver explanations like the built-in AI forecast explanation feature does.

How to eliminate wrong answers

Option A is wrong because Einstein Discovery is a separate tool for analyzing data patterns and generating stories, not for explaining a specific forecast's drivers; it would not directly show why the AI forecast differs from the manager's commit. Option B is wrong because disabling Einstein Forecasting eliminates the AI-driven insights entirely, which is an overreaction and does not help the manager understand the discrepancy. Option D is wrong because manually adjusting the AI forecast to match the commit defeats the purpose of using AI forecasting and introduces bias, undermining the model's objectivity.

Full explanation →
705
Multi-Selectmedium

A company is configuring Einstein Lead Scoring. Which TWO statements accurately describe how the feature works?

Select 2 answers
A.It assigns a score between 1 and 99 indicating conversion likelihood.
B.It only works with leads imported from external systems.
C.It scores opportunities based on deal size.
D.It can be used to prioritize leads in list views and reports.
E.It requires manual configuration of scoring rules by an admin.
AnswersA, D

Correct.

Why this answer

Option A is correct because Einstein Lead Scoring uses a predictive model to assign a score between 1 and 99 that reflects the likelihood a lead will convert to an opportunity. The score is calculated automatically based on historical conversion patterns and lead attributes, without requiring manual rule definition.

Exam trap

The trap here is that candidates often assume Einstein AI features require manual rule configuration (like traditional scoring tools), but Einstein Lead Scoring is fully automated and self-learning, making Option E a common distractor.

Full explanation →
706
MCQeasy

A Salesforce developer is building an AI model to predict customer churn. What is the most important ethical consideration when collecting training data?

A.Include as many features as possible
B.Focus on data from the last month for relevance
C.Collect only data that is necessary for the prediction
D.Use as much historical data as possible
AnswerC

Minimizing data collection protects customer privacy and reduces ethical risks.

Why this answer

The correct answer is B because data privacy is critical; collecting only necessary data minimizes risk. Option A is wrong because historical data can contain bias. Option C is wrong because older data may be less relevant.

Option D is wrong because more features can increase privacy risks and complexity.

Full explanation →
707
Multi-Selectmedium

A sales operations manager wants to use Einstein Opportunity Scoring to improve win rates. They want to view the opportunity score and understand why a particular score is high or low. Where can they see the score and explanation in Salesforce Lightning? (Choose TWO)

Select 2 answers
A.In Salesforce Mobile App under 'Today'
B.In the Activity Timeline
C.In the Opportunity list view as a column
D.In the Opportunity record's Einstein score component
E.In the Einstein Discovery dashboard
AnswersC, D

The score field can be added to list views for quick comparison.

Why this answer

Einstein Opportunity Scoring appears in the Opportunity record page as a score field and a component with explanation. It can also be added to list views and reports. The score field is automatically added to the Opportunity object.

Full explanation →
708
MCQmedium

What is the purpose of grounding in the Einstein Trust Layer?

A.To limit the AI's responses to pre-approved templates
B.To encrypt all data sent to the AI model
C.To ensure the AI only uses data from Salesforce, not third-party sources
D.To connect the AI to relevant CRM records so responses are accurate and context-aware
AnswerD

Grounding retrieves specific CRM data to inform the AI's output.

Why this answer

Grounding in the Einstein Trust Layer connects the AI model to relevant CRM records (e.g., Accounts, Opportunities, Cases) so that generated responses are based on accurate, up-to-date customer data rather than the model's general training data. This ensures context-aware and factual outputs while maintaining data privacy by not exposing raw records to the model.

Exam trap

Cisco often tests the misconception that grounding is about restricting data sources to Salesforce only, when in fact it is about connecting the AI to relevant CRM records (including integrated third-party data) to ensure accuracy and context.

How to eliminate wrong answers

Option A is wrong because grounding does not restrict responses to pre-approved templates; it dynamically retrieves CRM data to inform the AI's output, not force it into fixed formats. Option B is wrong because encryption (e.g., TLS for data in transit, AES-256 for data at rest) is a separate security layer within the Einstein Trust Layer, not the purpose of grounding. Option C is wrong because grounding can incorporate data from third-party sources if they are integrated into Salesforce (e.g., via External Objects or MuleSoft), not solely Salesforce-native data.

Full explanation →
709
MCQmedium

An admin wants to deploy an autonomous AI agent that can handle order cancellations end-to-end without human intervention. The agent needs to execute specific actions like querying order status and updating records. Which tool should they use?

A.Einstein Bots
B.Agentforce with Agent Builder
C.Einstein Next Best Action
D.Einstein GPT with Prompt Builder
AnswerB

Agentforce enables building autonomous agents that can perform actions independently based on topics.

Why this answer

Agentforce with Agent Builder is the correct tool because it is specifically designed to build autonomous AI agents that can execute end-to-end workflows, including querying order status and updating records, without human intervention. It uses a combination of large language models, deterministic actions, and guardrails to handle complex, multi-step tasks like order cancellations autonomously.

Exam trap

The trap here is that candidates confuse Einstein Bots (which require human-in-the-loop for actions) with autonomous agents, or they assume any Einstein AI tool with 'GPT' or 'Next Best Action' can execute backend transactions, when only Agent Builder provides the autonomous action execution capability.

How to eliminate wrong answers

Option A is wrong because Einstein Bots are designed for conversational, guided interactions with human customers, not for autonomous execution of backend actions like updating records without human oversight. Option C is wrong because Einstein Next Best Action provides recommendations for human agents or customers to act upon, not autonomous execution of actions. Option D is wrong because Einstein GPT with Prompt Builder generates text or content based on prompts, but it cannot autonomously execute API calls or database updates to complete an order cancellation workflow.

Full explanation →
710
MCQhard

A company uses Einstein's predictive lead scoring. The model inadvertently overweights leads from certain geographic regions. Which action aligns with Salesforce's Responsible AI principles?

A.Implement a feedback loop to continuously monitor and adjust.
B.Remove the geographic feature from the model.
C.Use the model as-is because it improves overall accuracy.
D.Only use the model for regions where it performs well.
AnswerA

Continuous monitoring allows ongoing bias detection and correction.

Why this answer

Option B is correct because continuous monitoring and adjustment is a key component of responsible AI. Option A is wrong because simply removing the geographic feature may not eliminate proxy variables. Option C is wrong because ignoring bias is unethical.

Option D is wrong because restricting usage may cause inequity.

Full explanation →
711
MCQmedium

A Service Cloud admin wants to deploy a chatbot that can handle common customer requests and escalate to a human agent when necessary. The chatbot must understand natural language variations. Which combination of tools should they use?

A.Einstein Case Classification with auto-response
B.Einstein Copilot with custom actions
C.Einstein Next Best Action with flows
D.Einstein Bots with intents and entities, and handoff to agent
AnswerD

Einstein Bots use NLP to understand intents and can escalate to live agents.

Why this answer

Einstein Bots in Service Cloud can be configured with intents and entities for NLP understanding and can hand off to human agents when needed.

Full explanation →
712
MCQeasy

A sales team wants to use Einstein Lead Scoring to prioritize leads. What is the primary benefit of using Einstein Lead Scoring over manual scoring?

A.It uses historical data to predict which leads are most likely to convert.
B.It automatically sends personalized emails to leads.
C.It ensures all leads are contacted within 24 hours.
D.It replaces the need for any manual lead qualification process.
AnswerA

Einstein Lead Scoring leverages machine learning on past lead conversions to assign a score.

Why this answer

Einstein Lead Scoring uses historical data and machine learning models to analyze patterns from past leads and their conversion outcomes. This allows it to assign a predictive score to each new lead, indicating the likelihood of conversion, which is more accurate and data-driven than manual scoring based on subjective criteria.

Exam trap

Salesforce often tests the misconception that AI features like Einstein Lead Scoring fully automate human tasks, when in reality they are designed to augment and prioritize, not replace, manual processes.

How to eliminate wrong answers

Option B is wrong because Einstein Lead Scoring does not automatically send emails; that function is handled by Einstein Engagement Scoring or automated email campaigns, not lead scoring. Option C is wrong because lead scoring prioritizes leads based on conversion likelihood, not on a time-based SLA like contacting within 24 hours. Option D is wrong because Einstein Lead Scoring augments, not replaces, manual qualification; human judgment is still needed for tasks like lead nurturing and complex decision-making.

Full explanation →
713
MCQhard

A financial institution deploys an AI system to recommend investment portfolios to retail clients. The system uses reinforcement learning to maximize returns based on client risk profiles. After six months, an internal audit reveals that the system has been consistently recommending high-risk, high-commission products to elderly clients with low risk tolerance, resulting in significant financial losses for those clients. The system's training data included historical transactions, which showed that elderly clients were less likely to complain or switch advisors. The institution's AI ethics policy mandates fairness, transparency, and accountability. The system currently provides no explanations for its recommendations, and there is no human oversight process. The compliance team needs to remediate the situation. Which course of action BEST addresses the ethical violations?

A.Disable the AI system and revert to manual portfolio management.
B.Add a disclaimer to all recommendations stating that past performance does not guarantee future results.
C.Adjust the model to lower the risk threshold for all clients.
D.Retrain the model on a balanced dataset, implement explainability features, and require human approval for high-risk recommendations to elderly clients.
AnswerD

This addresses bias, transparency, and accountability.

Why this answer

Option D is correct because it directly addresses the root cause of the ethical violations: biased training data (historical transactions where elderly clients were less likely to complain) and lack of transparency. Retraining on a balanced dataset mitigates the reinforcement learning model's exploitation of that bias, while explainability features (e.g., SHAP values or LIME) and human-in-the-loop approval for high-risk recommendations ensure accountability and fairness as mandated by the AI ethics policy.

Exam trap

Salesforce often tests the misconception that a single technical fix (like lowering risk thresholds or adding disclaimers) is sufficient to resolve ethical violations, when in fact a multi-pronged approach addressing data bias, transparency, and human oversight is required.

How to eliminate wrong answers

Option A is wrong because disabling the AI system and reverting to manual management is a reactive, non-scalable solution that does not address the underlying bias or provide a path to compliant AI deployment; it also ignores the potential benefits of AI when properly governed. Option B is wrong because adding a disclaimer does not fix the biased recommendations or lack of transparency; it merely shifts legal liability without correcting the model's unethical behavior or providing explanations. Option C is wrong because lowering the risk threshold for all clients is a blunt, one-size-fits-all approach that disregards individual risk profiles and may still result in inappropriate recommendations for elderly clients with low risk tolerance; it does not address the training data bias or the need for explainability and human oversight.

Full explanation →
714
Multi-Selecthard

Which TWO are best practices for mitigating bias in AI models when using Salesforce Einstein? (Choose two.)

Select 2 answers
A.Use diverse and representative training data that reflects the target population.
B.Exclude demographic features that have small sample sizes to avoid statistical noise.
C.Prioritize accuracy for the majority group to maximize overall performance.
D.Rely on convenience sampling to quickly gather a large dataset.
E.Conduct regular bias audits using Einstein's fairness evaluation tools.
AnswersA, E

Diverse data reduces the risk of biased outcomes.

Why this answer

Options A and D are correct. Option B is wrong because convenience sampling can introduce bias. Option C is wrong because ignoring small groups can perpetuate bias.

Option E is wrong because focusing only on high-accuracy groups may sacrifice fairness.

Full explanation →
715
MCQhard

A financial services firm uses Data Cloud to enrich sales data with external credit scores via an API. They set up a Data Action to call the credit bureau API for each new lead. Over time, API costs are rising, and the action is slowing down lead processing. They only need credit scores for leads with a high probability of conversion. What is the best approach to reduce costs and improve performance?

A.Remove the Data Action and manually verify credit scores for top leads
B.Apply a Data Transform to filter leads that have incomplete data before the Data Action
C.Schedule the Data Action to run daily in batch instead of real-time
D.Use a Calculated Insight to score leads based on internal data and only invoke the Data Action for leads with a high probability of conversion
AnswerD

Selectively calls API only for promising leads, reducing costs and load.

Why this answer

Option C is the best solution because it uses a Calculated Insight to compute a conversion probability score based on internal data, then only triggers the Data Action for leads above a threshold. This reduces API calls significantly. Option A filters but still requires an initial call? Actually, a Data Transform filter could be applied before the Data Action, but a Calculated Insight allows dynamic scoring.

Option B runs the action in batch, but still calls for all leads. Option D removes the action entirely, losing the enrichment.

Full explanation →
716
MCQhard

A data scientist is building a custom AI model using Salesforce Data Cloud to predict customer churn. They want to ensure that the model does not inadvertently use gender as a feature to avoid biased predictions. Which step is MOST appropriate?

A.Exclude gender from the feature set used for model training
B.Use gender as a feature but ignore the model predictions for certain groups
C.Allow gender in training but use a post-processing technique to adjust scores
D.Include gender as a feature and then apply a fairness constraint during training
AnswerA

Excluding protected attributes is a direct way to prevent the model from using them; it aligns with data minimisation.

Why this answer

The best practice is to exclude protected attributes from the model features. Data minimisation supports this. Auditing for bias is also important but does not prevent the model from using the attribute in the first place.

Full explanation →
717
MCQmedium

An organization wants to build an autonomous AI agent that can handle customer inquiries order status and return requests without human intervention. Which Salesforce tool allows them to build such an agent with topics and actions?

A.Agentforce
B.Einstein Copilot
C.Einstein Bots
D.Einstein Next Best Action
AnswerA

Correct. Agentforce with Agent Builder enables creation of autonomous agents with topics and actions.

Why this answer

Agentforce is the correct answer because it is the Salesforce tool specifically designed to build autonomous AI agents that can handle customer inquiries, order status, and return requests without human intervention. It allows you to define topics (e.g., 'Order Status') and actions (e.g., 'Lookup Order') that the agent can execute independently, using natural language processing and integration with Salesforce data.

Exam trap

The trap here is that candidates often confuse Einstein Copilot (a user-facing assistant) with Agentforce (a builder for autonomous agents), or they assume Einstein Bots can handle complex autonomous tasks when they are actually limited to simpler, scripted interactions.

How to eliminate wrong answers

Option B is wrong because Einstein Copilot is a conversational AI assistant that helps users interact with Salesforce data and workflows, but it is not designed to build autonomous agents with topics and actions; it is more of a copilot for users. Option C is wrong because Einstein Bots are rule-based or AI-powered chatbots that can handle simple inquiries, but they lack the autonomous agent capabilities and the structured topics/actions framework that Agentforce provides for complex, multi-step tasks. Option D is wrong because Einstein Next Best Action is a recommendation engine that suggests the next best action for a user (e.g., a sales rep) based on AI models, not an autonomous agent that handles customer inquiries end-to-end.

Full explanation →
718
MCQeasy

When implementing AI in Salesforce, which practice best supports the ethical principle of transparency?

A.Provide human-readable explanations for each AI prediction
B.Use proprietary algorithms without disclosing their logic
C.Deploy a complex neural network model without interpretability features
D.Only report overall model accuracy metrics to end users
AnswerA

Explanations enable understanding and trust.

Why this answer

Transparency requires that the logic and outcomes of AI systems are understandable. Option A is correct because providing explanations for predictions allows users to understand and trust the AI. Option B (keeping proprietary algorithms secret) hinders transparency.

Option C (using complex models without explanation) obscures decision-making. Option D (only reporting accuracy metrics) does not explain specific decisions.

Full explanation →
719
Multi-Selecteasy

Which THREE of the following are Einstein features available for Sales Cloud? (Choose three.)

Select 3 answers
A.Einstein Lead Scoring
B.Einstein Opportunity Scoring
C.Einstein Activity Capture
D.Einstein Bot
E.Einstein Analytics
AnswersA, B, C

Predicts lead conversion.

Why this answer

Einstein Lead Scoring is a correct Einstein feature for Sales Cloud because it uses predictive AI models to automatically score leads based on historical conversion data, helping sales reps prioritize high-quality leads. It is natively integrated into Sales Cloud without requiring additional licenses or complex setup, leveraging standard Salesforce objects and fields.

Exam trap

Salesforce often tests candidates' ability to distinguish between Einstein features that are native to Sales Cloud (like scoring and activity capture) versus those that are cross-cloud or require separate licenses (like Einstein Bot or Einstein Analytics), leading to confusion when options include features that are technically available but not part of the core Sales Cloud Einstein set.

Full explanation →
720
MCQmedium

A financial services company uses Einstein AI to recommend credit limits. The model tends to assign lower limits to applicants from a certain region. Which action best aligns with ethical AI practices?

A.Replace the AI model with a simpler rule-based system
B.Investigate the data and model for bias, and adjust the model if necessary
C.Manually increase credit limits for all applicants from that region
D.Ignore the pattern since the model is statistically valid
AnswerB

Investigation and adjustment is the correct ethical approach.

Why this answer

Option A (Investigate the data and model for bias, and adjust the model if necessary) is correct because ethical AI requires proactive identification and mitigation of bias. Option B (ignoring the pattern) could allow discrimination. Option C (increasing limits for that region without analysis) may not be justified.

Option D (using a different AI model without investigation) does not address the root cause.

Full explanation →
721
MCQmedium

A data engineer needs to create a feature that represents the average purchase amount per customer over the last 30 days. The transactional data is timestamped. Which feature engineering technique is most appropriate?

A.Sum of all purchase amounts per customer
B.Rolling average of purchase amounts over a 30-day window
C.Count of purchases per customer
D.Minimum purchase amount per customer
AnswerB

Rolling average matches the requirement.

Why this answer

Option B is correct because a rolling average over a 30-day window directly computes the average purchase amount per customer for only the most recent 30 days of transactions, which matches the requirement of a time-sensitive feature. This technique uses a sliding window function (e.g., AVG() with a ROWS or RANGE frame in SQL, or rolling().mean() in pandas) that respects the timestamp order, ensuring only relevant data contributes to the feature.

Exam trap

Salesforce often tests the distinction between simple aggregation (like sum or count) and time-windowed aggregation, trapping candidates who overlook the 'over the last 30 days' temporal constraint and choose a static aggregate instead.

How to eliminate wrong answers

Option A is wrong because summing all purchase amounts per customer ignores the 30-day time constraint and would include historical data outside the window, producing a feature that does not reflect recent behavior. Option C is wrong because counting purchases per customer measures frequency, not the average amount spent, and also lacks the time window restriction. Option D is wrong because the minimum purchase amount per customer is a different aggregate (minimum) that does not capture the central tendency of spending and similarly ignores the 30-day window.

Full explanation →
722
MCQhard

A company uses Einstein GPT for Sales to generate personalized email drafts. They want to ensure that the generated emails consistently include the recipient's company name and a specific discount offer from the opportunity. Which configuration is required?

A.Create a prompt template in Prompt Builder with merge fields for Company and Discount
B.Configure Einstein Email Insights to highlight opportunities with discount information
C.Set up a Flow in Einstein Next Best Action to inject data into the email
D.Use Einstein Recommendation Builder to define dynamic email components
AnswerA

Prompt Builder enables merging of Salesforce field values into generative AI prompts, ensuring consistency.

Why this answer

Option A is correct because Prompt Builder allows admins to create prompt templates that include merge fields for dynamic data, such as the recipient's company name and a discount offer from the opportunity. When Einstein GPT for Sales generates email drafts, it uses these merge fields to pull the specific values from the Salesforce record, ensuring consistency across all generated emails.

Exam trap

The trap here is that candidates may confuse Einstein GPT's content generation capabilities with other Einstein features like Next Best Action or Recommendation Builder, which are designed for recommendations and actions rather than direct, merge-field-driven text generation.

How to eliminate wrong answers

Option B is wrong because Einstein Email Insights is an analytics tool that provides visibility into email engagement metrics (e.g., open rates, click rates) and does not have the capability to inject or control dynamic content like company names or discount offers into email drafts. Option C is wrong because Einstein Next Best Action is designed to recommend the next best action (e.g., a prompt or offer) to users based on context, but it does not directly generate or populate email content with merge fields; it relies on flows or recommendations, not prompt templates. Option D is wrong because Einstein Recommendation Builder is used to create product or content recommendations (e.g., 'Customers also bought') for websites or emails, not to define dynamic text components with merge fields for personalized email drafts.

Full explanation →
723
MCQmedium

A company wants to build a custom AI model that predicts whether a customer will churn within the next 30 days, using data from multiple Salesforce objects. The prediction should output a score from 0 to 100. Which Einstein feature is most appropriate?

A.Einstein Discovery
B.Einstein Case Classification
C.Einstein Prediction Builder
D.Einstein Next Best Action
AnswerC

Prediction Builder enables creating a custom binary classification model predicting churn, with a score field.

Why this answer

Einstein Prediction Builder is the correct choice because it allows users to build custom predictive models using data from multiple Salesforce objects without writing code, and it outputs a score (0–100) representing the likelihood of a specific outcome, such as customer churn within 30 days. This feature is designed for point-and-click creation of binary classification models that generate a probability score, directly matching the requirement.

Exam trap

The trap here is that candidates confuse Einstein Discovery (which provides insights and explanations) with Einstein Prediction Builder (which outputs a custom predictive score), because both use AI and can analyze data, but only Prediction Builder generates a 0–100 probability score for a user-defined outcome.

How to eliminate wrong answers

Option A is wrong because Einstein Discovery is an automated analytics and insights tool that identifies patterns and correlations in data but does not output a custom predictive score for a specific binary outcome like churn; it provides explanations and recommendations, not a 0–100 prediction score. Option B is wrong because Einstein Case Classification is specifically designed to automatically classify and route support cases based on intent or topic, not to predict customer churn using data from multiple Salesforce objects. Option D is wrong because Einstein Next Best Action recommends the next best action to take for a customer based on predefined rules or AI models, but it does not build a custom predictive model that outputs a churn score; it consumes predictions from other tools.

Full explanation →
724
MCQeasy

A company uses an AI model to screen job applicants. They discover the model is rejecting candidates from a certain demographic at a higher rate. Which ethical principle is most clearly violated?

A.Fairness
B.Transparency
C.Privacy
D.Accountability
AnswerA

Correct. The model's bias against a demographic violates fairness.

Why this answer

Fairness requires that AI systems do not discriminate against groups. The model's disparate impact violates fairness.

Full explanation →
725
MCQmedium

A university uses an AI system to predict first-year student retention. The system uses factors such as high school GPA, SAT scores, and socioeconomic indicators. After two years, administrators notice that the model consistently predicts lower retention probabilities for students from low-income families, even when their academic profiles are strong. The university's mission emphasizes equity and inclusion. The admissions office is considering using the predictions to allocate support resources. The model's accuracy on historical data is 85%. What should the university do to align with ethical AI principles?

A.Remove socioeconomic indicators from the model inputs.
B.Abandon the AI system and rely on human advisors for resource allocation.
C.Retrain the model with a fairness constraint such as equalized odds to reduce income-based disparities.
D.Use the predictions as-is because they are accurate for the majority of students.
AnswerC

Fairness constraints directly address the ethical concern while maintaining predictive power.

Why this answer

Option B is correct because implementing a fairness metric like equal opportunity ensures the model does not disadvantage a protected group. Option A would perpetuate inequity. Option C removes a feature that may still allow bias through correlated variables.

Option D abandons a potentially useful tool without addressing bias.

Full explanation →
726
MCQhard

A Salesforce admin wants to use Einstein GPT to automatically generate a meeting follow-up email after a sales rep closes a meeting record in Salesforce. They want the email to include the meeting summary, key action items, and next steps. Which combination of features should they use?

A.Einstein Discovery and Prompt Builder
B.Prompt Builder with a Sales Email template and Einstein Activity Capture
C.Einstein Conversation Insights and Einstein GPT for Sales
D.Einstein Activity Capture and Einstein Bots
AnswerB

Activity Capture logs the meeting, and Prompt Builder generates the email from meeting data.

Why this answer

Prompt Builder can create a Sales Email prompt template that pulls data from the meeting record. Einstein Activity Capture can log the meeting, but generation is done via Prompt Builder. Einstein Bots and Discovery are not relevant.

Full explanation →
727
MCQmedium

Refer to the exhibit. A Salesforce admin configured the Einstein Trust Layer policy shown. What is the effect of this policy on AI model usage?

A.All fields in the org will be masked to protect customer privacy.
B.AI models will not be able to use the configured fields, and model insights are disabled.
C.AI models can still use the fields but feature importance insights are blocked.
D.AI models receive masked data for those fields, but feature importance insights are still available.
AnswerD

Masking hides actual values; insights are independent.

Why this answer

The Einstein Trust Layer policy configured to mask specific fields ensures that sensitive data is replaced with masked values before being sent to the AI model. This preserves data privacy while still allowing the model to generate predictions and insights. Feature importance insights remain available because they are computed from the masked data, not the original values.

Exam trap

The trap here is that candidates often assume masking blocks all AI functionality, but feature importance insights are still available because they rely on patterns in the masked data, not the original values.

How to eliminate wrong answers

Option A is wrong because the policy only masks the configured fields, not all fields in the org. Option B is wrong because masking does not disable model insights; the AI model can still use the masked data to generate predictions and insights. Option C is wrong because feature importance insights are not blocked; they are still computed and available even when fields are masked.

Full explanation →
728
Multi-Selecthard

An admin is troubleshooting why Einstein Lead Scoring is not generating scores for some leads. They have enabled the feature and assigned permission sets. Which two factors could cause scores to be missing?

Select 2 answers
A.The lead's status is 'Converted' and scoring is configured to exclude converted leads
B.The lead has a custom field that is not included in the model
C.The lead does not have an email address
D.The lead's owner does not have a Salesforce license
E.There are fewer than 100 open leads in the org
AnswersA, E

Converted leads may be excluded by default or admin settings.

Why this answer

Option A is correct because Einstein Lead Scoring can be configured to exclude leads with a 'Converted' status. If the scoring model is set to skip converted leads, those leads will not receive a score, even if the feature is enabled and permission sets are assigned. This is a common configuration setting that directly prevents scoring for converted records.

Exam trap

The trap here is that candidates may assume missing scores are due to data quality issues (like missing email) or licensing, when the actual cause is a deliberate configuration setting that excludes converted leads from scoring.

Full explanation →
729
MCQmedium

A service agent needs quick access to relevant knowledge articles while handling a case. Which Einstein feature can suggest the most relevant articles automatically?

A.Einstein Next Best Action
B.Einstein Case Classification
C.Einstein Article Recommendations
D.Einstein Service GPT
AnswerC

Article Recommendations uses AI to suggest relevant knowledge articles to agents.

Why this answer

Einstein Article Recommendations (C) is the correct feature because it uses AI to automatically surface the most relevant knowledge articles based on the context of the case, such as subject, description, and product. This directly addresses the need for quick access to relevant articles without manual search, leveraging Salesforce's predictive AI to match case data with article content.

Exam trap

The trap here is that candidates often confuse Einstein Article Recommendations with Einstein Next Best Action, because both involve 'recommendations,' but Next Best Action is for customer-facing offers or actions, not for agent-facing knowledge retrieval.

How to eliminate wrong answers

Option A is wrong because Einstein Next Best Action is designed to recommend the next best action or offer to a customer based on real-time data and business rules, not to suggest knowledge articles for agent use. Option B is wrong because Einstein Case Classification automatically categorizes and routes cases to the appropriate queue or agent, but it does not provide article recommendations. Option D is wrong because Einstein Service GPT is a generative AI tool that can draft responses or summarize cases, but it is not specifically built for suggesting relevant knowledge articles; its primary function is content generation, not retrieval of existing articles.

Full explanation →
730
MCQmedium

A sales manager wants to predict which deals are likely to close this quarter. The CRM has rich historical data on won/lost opportunities, deal amount, and sales stage. Which AI approach is best suited for this task?

A.Predictive AI for opportunity scoring
B.Generative AI to create new sales content
C.Unsupervised learning to cluster opportunities
D.Reinforcement learning for sales strategy
AnswerA

Predictive AI can use historical data to predict the likelihood of a deal closing.

Why this answer

Predictive AI uses historical data to forecast outcomes, making it ideal for opportunity scoring.

Full explanation →
731
Multi-Selecthard

A Salesforce admin is creating an Agentforce agent in Agent Builder to handle order status inquiries. The agent needs to look up order data from a custom object 'Order__c' and respond with the status. Which THREE components must the admin configure in Agent Builder?

Select 3 answers
A.Prompt instructions that tell the agent how to respond
B.Intents and entities for natural language understanding
C.An action that queries the Order__c object
D.A topic named 'Order Status'
E.A Flow to route the conversation to a human agent
AnswersA, C, D

Prompt instructions define the agent's tone, format, and behavior.

Why this answer

Topics group related conversations, actions perform data operations, and prompt instructions guide the agent's behavior. Intents are for Einstein Bots, not Agent Builder. Flows are used within actions, but the action itself is required.

Full explanation →
732
MCQmedium

A marketing team wants to recommend products to customers in an Experience Cloud community using AI. Which feature should they implement?

A.Einstein GPT for Sales
B.Einstein Article Recommendations
C.Einstein Next Best Action
D.Einstein Recommendation Builder
AnswerD

Recommendation Builder is designed for Experience Cloud to recommend products or content based on AI.

Why this answer

Einstein Recommendation Builder is the correct feature because it allows marketers to create and deploy AI-powered product recommendations specifically within Experience Cloud communities, using customer behavior and profile data to personalize the community experience. Unlike other Einstein features, Recommendation Builder is designed for community sites and can be configured without code to recommend products, articles, or custom records.

Exam trap

The trap here is that candidates often confuse Einstein Next Best Action (which is broader and action-oriented) with product recommendations, but Next Best Action is rule-based and not optimized for the specific use case of recommending products in a community without custom development.

How to eliminate wrong answers

Option A is wrong because Einstein GPT for Sales is a conversational AI tool for sales teams to generate emails, call summaries, and deal insights, not a product recommendation engine for Experience Cloud communities. Option B is wrong because Einstein Article Recommendations is focused on suggesting knowledge articles (e.g., help docs) within communities or service consoles, not products for marketing purposes. Option C is wrong because Einstein Next Best Action is a decision engine that presents the most relevant actions (e.g., offers, tasks) to agents or customers based on rules and AI, but it is not specifically designed for product recommendations in a community context and requires more complex configuration than Recommendation Builder.

Full explanation →
733
MCQmedium

Refer to the exhibit. Which ethical principle is most at risk based on this AI governance configuration?

A.Transparency
B.Fairness
C.Privacy
D.Accountability
AnswerD

Lack of human override and explanation reduces accountability.

Why this answer

The configuration lacks human override and does not require explanations, which undermines accountability. Option C (Accountability) is correct. Option A (Fairness) is not directly addressed but audit trail may help.

Option B (Privacy) is not mentioned. Option D (Transparency) is partially addressed by audit trail but not explanation. The absence of human override is a key accountability gap.

Full explanation →
734
MCQhard

A company uses an NLP model to detect customer intent from chat messages. The model correctly identifies 'billing question' 90% of the time for actual billing questions, but also flags many non-billing messages as billing (false positives). Which metric should the team prioritize to reduce false alarms?

A.Accuracy
B.F1 score
C.Precision
D.Recall
AnswerC

Correct. Precision = TP/(TP+FP); higher precision means fewer false alarms.

Why this answer

Precision focuses on the proportion of positive identifications that are correct; improving precision reduces false positives.

Full explanation →
735
MCQmedium

An e-commerce company uses AI to provide product recommendations. The model suggests popular items but fails to personalize for individual users. Which type of learning could improve personalization?

A.Generative AI
B.Unsupervised learning
C.Reinforcement learning
D.Supervised learning
AnswerD

Supervised learning can train on user-item interactions to predict personalized recommendations.

Why this answer

Supervised learning can use user purchase history as labels to predict what a specific user might buy, enabling personalization.

Full explanation →
736
MCQeasy

A company is preparing data for Einstein Article Recommendation. Which data source is most appropriate for training the model?

A.Historical article view and click data.
B.Org metadata.
C.System debug logs.
D.User profile data only.
AnswerA

This captures user preferences directly.

Why this answer

Einstein Article Recommendation uses supervised machine learning to predict which articles users are likely to find relevant. The model must be trained on historical user engagement signals—specifically article view and click data—to learn patterns of relevance. Without this behavioral data, the model cannot establish a correlation between user actions and article content.

Exam trap

Salesforce often tests the misconception that static data like user profiles or org metadata can substitute for behavioral training data, but the model fundamentally requires historical interaction signals to learn relevance.

How to eliminate wrong answers

Option B is wrong because org metadata (e.g., company name, industry) provides only static contextual information and lacks the user-article interaction signals required for training a recommendation model. Option C is wrong because system debug logs contain low-level technical events (e.g., errors, stack traces) that are irrelevant to user content preferences and would introduce noise rather than meaningful training features. Option D is wrong because user profile data alone (e.g., role, department) does not capture which articles users actually viewed or clicked, so the model cannot learn relevance from user behavior.

Full explanation →
737
MCQeasy

What does the term 'hallucination' refer to in the context of generative AI?

A.The model only works when the user prompts it with specific keywords
B.The model has a bug in the training algorithm
C.The model generates creative but accurate content
D.The model produces outputs that are not grounded in the training data or reality
AnswerD

Correct: hallucination refers to fabricated or incorrect information.

Why this answer

Hallucination is when a generative AI model produces content that is factually incorrect or nonsensical but presented as if it were true.

Full explanation →
738
MCQmedium

A CRM team wants to predict the expected revenue from each opportunity. The data includes opportunity amount, close date, stage, and historical win rates. Which type of AI is best suited?

A.Unsupervised learning
B.Predictive AI
C.Generative AI
D.Reinforcement learning
AnswerB

Correct. Predictive AI forecasts numeric outcomes using regression models.

Why this answer

Predictive AI uses historical data to forecast outcomes. Regression (a type of predictive AI) predicts a continuous value like revenue.

Full explanation →
739
MCQhard

An administrator is setting up an Einstein Bot for a service cloud. The bot needs to understand when a customer types 'cancel order' to route them to a cancellation flow. Which bot component should the administrator configure to recognize this phrase?

A.Intent
B.Dialog
C.Action
D.Entity
AnswerA

Correct. Intents are used to classify user input into categories like 'cancel order'.

Why this answer

An intent represents the goal or purpose behind a user's input, such as 'cancel order.' In Einstein Bots, intents are trained to recognize specific phrases and map them to corresponding dialog flows. By configuring an intent for 'cancel order,' the bot can accurately route the customer to the cancellation flow without relying on exact keyword matching.

Exam trap

The trap here is confusing 'intent' with 'entity' — candidates often think extracting the phrase 'cancel order' is an entity task, but entities capture data values (e.g., order ID), not the action or goal expressed by the user.

How to eliminate wrong answers

Option B is wrong because a dialog defines the conversation path and responses after an intent is recognized, not the recognition of the phrase itself. Option C is wrong because an action performs a specific task (e.g., calling an API or updating a record) after the intent is identified, not the initial phrase recognition. Option D is wrong because an entity extracts specific data from the user's input (e.g., order number), not the overall intent or purpose of the phrase.

Full explanation →
740
MCQhard

A retail company deploys an AI system that adjusts prices dynamically based on customer browsing history. The system charges higher prices to returning customers. This practice is known as:

A.Lack of transparency
B.Unfair price discrimination
C.Personalization
D.Dynamic pricing
AnswerB

Charging loyal customers more is unfair and unethical.

Why this answer

Option D is correct because price discrimination based on customer data is unfair and often unethical. Option A is wrong because personalization is about tailoring experience, not price gouging. Option B is wrong as dynamic pricing is legal but can be unethical.

Option C is wrong because transparency issue is secondary.

Full explanation →
741
MCQmedium

An admin wants to generate a summary of a sales call recording automatically. Which Einstein feature should be used?

A.Einstein Activity Capture
B.Einstein Conversation Insights
C.Einstein GPT for Sales
D.Einstein Email Insights
AnswerB

Why this answer

Einstein Conversation Insights is the correct feature because it is specifically designed to analyze sales call recordings and generate summaries, key moments, and action items using natural language processing (NLP) and speech-to-text technology. It ingests audio from calls, transcribes them, and applies AI to extract insights such as customer sentiment, competitor mentions, and next steps, directly meeting the admin's requirement.

Exam trap

The trap here is that candidates often confuse Einstein Conversation Insights with Einstein Activity Capture, assuming that any feature with 'Activity' in the name can handle call recordings, when in fact Activity Capture is limited to email and calendar sync and lacks audio processing capabilities.

How to eliminate wrong answers

Option A is wrong because Einstein Activity Capture syncs emails, events, and contacts between Salesforce and external systems (e.g., Outlook, Gmail) but does not process or summarize call recordings. Option C is wrong because Einstein GPT for Sales generates personalized content like email drafts or call scripts using generative AI, but it does not analyze existing call recordings to produce summaries. Option D is wrong because Einstein Email Insights analyzes email metadata and engagement patterns (e.g., open rates, reply times) to prioritize leads, not audio or transcript data from sales calls.

Full explanation →
742
MCQmedium

A company uses Einstein Lead Scoring but notices that many leads with high scores are not converting. The admin wants to understand which factors are influencing the score. Where can the admin find this information?

A.In the Einstein Setup menu under Lead Scoring Models
B.By contacting Salesforce support for a model explanation
C.On the lead record, in the Einstein Lead Score component, by clicking 'Score Factors'
D.By running a report on the Lead object with the Einstein Score field
AnswerC

The lead record displays the score and an expandable 'Score Factors' section showing contributing fields.

Why this answer

Option C is correct because the Einstein Lead Score component on the lead record includes a 'Score Factors' link that, when clicked, displays the top positive and negative factors influencing the lead score. This is the designated UI element for admins to gain transparency into which attributes (e.g., source, industry, behavior) are driving the score, without needing to access setup menus or external support.

Exam trap

The trap here is that candidates assume the model configuration or a report would show factor details, but Salesforce deliberately hides factor breakdowns behind the lead record component to emphasize that this is a per-record, UI-level insight, not a setup or reporting feature.

How to eliminate wrong answers

Option A is wrong because the Einstein Setup menu under Lead Scoring Models is used to configure scoring models (e.g., select fields, set model status), not to view per-lead factor breakdowns. Option B is wrong because Salesforce support does not provide model explanations for standard Einstein features; the factor details are self-service via the lead record. Option D is wrong because a report on the Lead object with the Einstein Score field shows only the numeric score, not the contributing factors; factor details are not exposed in reportable fields.

Full explanation →
743
MCQhard

An AI Associate reviews the Lead Scoring model exhibit. What is the primary ethical concern with this model?

A.The model uses too many features.
B.The model has low recall, potentially missing minority class leads.
C.The model is not explainable.
D.The training data is imbalanced.
AnswerB

Low recall can lead to underrepresentation of certain groups.

Why this answer

The primary ethical concern is that the model has low recall, meaning it fails to identify a significant portion of actual positive leads (the minority class). In a lead scoring context, this can result in missed business opportunities and potential bias against certain customer segments, as the model systematically overlooks valuable leads that do not fit the majority pattern.

Exam trap

Salesforce often tests the distinction between a technical problem (like imbalanced data) and its ethical consequence (like low recall causing unfair outcomes), so candidates mistakenly pick the technical cause (D) instead of the ethical impact (B).

How to eliminate wrong answers

Option A is wrong because using many features is not inherently an ethical concern; feature selection impacts performance and overfitting, but the ethical issue here is about fairness and missed opportunities, not feature count. Option C is wrong because the exhibit does not indicate a lack of explainability; the model could be a decision tree or logistic regression that is inherently interpretable, and explainability is not the primary ethical issue raised by the confusion matrix. Option D is wrong because imbalanced training data is a technical challenge that can lead to low recall, but the primary ethical concern is the consequence of that imbalance—specifically the model's low recall causing minority class leads to be missed—not the imbalance itself.

Full explanation →
744
Multi-Selecteasy

A sales manager wants to use Einstein Lead Scoring but is concerned about transparency for the sales team. Which TWO features should they enable to provide explainability? (Choose 2)

Select 2 answers
A.Einstein Copilot
B.Score Factors on the lead record
C.Einstein Activity Capture
D.A custom field indicating the score is AI-generated
E.Einstein Trust Layer audit trail
AnswersB, D

Score Factors display the top contributing fields and their impact.

Why this answer

Score Factors show why a lead scored as it did, and labeling AI-generated scores helps reps understand that the score is AI-driven.

Full explanation →
745
MCQhard

A data scientist trains a churn prediction model on CRM data that includes customer tenure, support ticket count, and last purchase date. The model achieves 95% accuracy on training data but only 60% on a holdout validation set. What is the most likely issue?

A.The validation set contains data leakage
B.The model needs more features
C.The model is overfitting the training data
D.The model is underfitting the training data
AnswerC

Correct: high training accuracy but much lower validation accuracy indicates overfitting.

Why this answer

Large gap between training accuracy and validation accuracy is a classic sign of overfitting. The model memorized training data and fails to generalize.

Full explanation →
746
MCQmedium

A company deploys Einstein Recommendation Builder on its e-commerce site. The recommendations are not personalized. What is the most likely cause?

A.The model has not been trained with enough user behavior data.
B.The company did not hire a data scientist to tune the model.
C.The recommendation engine is not syncing in real-time with the website.
D.The product catalog is too large for the model to process.
AnswerA

Personalization requires sufficient historical data.

Why this answer

Einstein Recommendation Builder relies on user interaction data to personalize. If insufficient data exists, recommendations become generic. Option A is correct.

Option B is wrong because real-time sync is not required. Option C is wrong because the builder can work without a data scientist. Option D is wrong because the model can recommend products beyond categories.

Full explanation →
747
MCQeasy

A retail company uses an AI recommendation engine to suggest products to online shoppers. The engine uses past purchase history and browsing behavior. Recently, a customer advocacy group publishes a report showing that the engine recommends higher-priced products to customers in affluent zip codes and lower-priced products to customers in lower-income areas, even when both groups have similar browsing histories. The company's revenue has increased since implementing the engine, and marketing teams are pleased. However, the company wants to maintain a reputation for fairness. Which action should the company take?

A.Show only the most popular products to all customers regardless of browsing history.
B.Discontinue the AI recommendation system entirely.
C.Keep the current system as it maximizes revenue.
D.Implement fairness constraints to ensure similar recommendation distributions across demographic groups.
AnswerD

Fairness constraints balance business goals with ethical considerations, maintaining personalization while avoiding bias.

Why this answer

Option B is correct because implementing fairness constraints ensures recommendations are not systematically skewed, while still allowing personalization. Option A prioritizes revenue over ethics. Option C is too drastic and would lose benefits.

Option D reduces personalization and may not be effective.

Full explanation →
748
MCQhard

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?

A.Remove the zip code feature from the model inputs.
B.Retrain the model with a balanced dataset that includes more examples from underrepresented neighborhoods and enforce fairness constraints.
C.Adjust the approval threshold lower only for applicants from low-income neighborhoods.
D.Continue using the existing model since it has high accuracy and reduces defaults.
AnswerB

Balanced data reduces bias and fairness constraints ensure equitable treatment, aligning with ethical AI principles.

Why this answer

Option B is correct because retraining with balanced data mitigates the representation bias, addressing the root cause. Option A ignores the fairness issue. Option C removes a feature that may be a proxy for other factors, but it may not eliminate bias if other correlated features remain.

Option D adjusts thresholds only for some groups, which could be considered unfair and may not be accepted by regulators.

Full explanation →
749
MCQhard

A company uses Einstein Conversation Insights to analyze sales calls. They want to identify when a competitor is mentioned and automatically log that mention to the opportunity. Which feature should they configure?

A.Einstein Next Best Action
B.Einstein Email Insights
C.Einstein Activity Capture
D.Keyword Tracking in Conversation Insights
AnswerD

Keyword tracking captures specific terms and can trigger actions such as logging.

Why this answer

Einstein Conversation Insights allows keyword tracking, where you can define keywords (e.g., competitor names) and set actions like logging to a record.

Full explanation →
750
MCQeasy

Refer to the exhibit. Which ethical principle is most directly violated?

A.Privacy
B.Transparency
C.Accountability
D.Fairness
AnswerA

PII leakage violates privacy.

Why this answer

Option B is correct because PII leakage violates privacy. Option A is wrong fairness is about bias. Option C is wrong transparency is about openness.

Option D is wrong accountability is about responsibility, but the immediate violation is privacy.

Full explanation →
Page 9

Page 10 of 14

Page 11

All pages

1234567891011121314

Practice AI Associate by domain

Target a specific domain to shore up weak areas.

Ethical AI and Data PrivacySalesforce Einstein AI FeaturesAI FundamentalsAI Capabilities in CRMEthical Considerations of AIData for AI
See all domains with question counts →