AI Associate Practice Exam 51 — 20 Questions

Question 1

A regional bank is deploying an Einstein Discovery model that predicts which mortgage applications are likely to default. The compliance team asks the AI team to describe how the model makes its decisions so they can document the logic for regulators. Which characteristic of the model is the compliance team asking about?

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

Accepted Answer

Model explainability. Explainability lets stakeholders see which input factors drove a specific prediction, such as why a mortgage application was scored as likely to default. Einstein Discovery provides model insights, top predictors, and reason codes that make this reasoning visible. For regulated lending decisions, this transparency is what allows the bank to document and defend the model's logic during audits.

Answer

Data completeness

Answer

Data storage cost

Answer

Model training speed

Question 2

A retail company is evaluating whether to build its own machine learning model or adopt an out-of-the-box Salesforce predictive feature. The data science lead explains that the chosen approach must learn patterns from labeled historical examples, then apply those patterns to new, unseen records to output a prediction. Which term best describes this category of AI capability?

Accepted Answer

Supervised learning. Learning from historical records that contain known outcomes and then predicting those outcomes for new records is the definition of supervised learning. Unsupervised learning works without labels, reinforcement learning learns from reward feedback through actions, and generative AI produces new content, so none of them fits the described labeled-prediction workflow.

Answer

Reinforcement learning

Answer

Generative AI

Answer

Unsupervised learning

Question 3

A sales operations team wants to ensure that reps see a prioritized list of opportunities that are most likely to close, based on historical patterns in their Salesforce opportunity data. They have Einstein Opportunity Scoring enabled. Where should they surface these scores so reps can act on them directly within their daily workflow?

Accepted Answer

On the Opportunity list view as a sortable column and on the Opportunity record detail page as a field.. Einstein Opportunity Scoring writes a predictive score to the opportunity record. To make it actionable, the score field should be added to list views for sorting and to the record page for visibility. This allows reps to focus on high-probability deals within their existing workflow, which is the primary purpose of the feature.

Answer

In a custom report built with CRM Analytics, refreshed daily, and distributed via email.

Answer

As a Chatter post automatically generated when the score changes by more than 10 points.

Answer

In a custom Lightning component that queries the Einstein Discovery API in real time for each opportunity.

Question 4

A Salesforce administrator is evaluating Einstein Discovery to help a retail company predict which customers are likely to churn. The admin wants to explain to stakeholders how Einstein Discovery generates its predictions. Which statement accurately describes a core capability of Einstein Discovery?

Accepted Answer

It automatically explores datasets, builds multiple models, and surfaces statistically significant drivers of the predicted outcome with explanations.. Einstein Discovery automates dataset exploration, model creation, and statistical analysis, then explains which variables drive the predicted outcome. For a churn use case, this means the retail company gets not just a churn probability but also the key drivers, such as declining purchase frequency or unresolved cases, which stakeholders can act on directly in Salesforce.

Answer

It stores predictions in an external data warehouse and requires a separate BI tool to visualize results.

Answer

It only produces a single linear regression model and cannot handle non-linear relationships in the data.

Answer

It requires data scientists to manually write Python notebooks that are then uploaded to Salesforce for execution.

Question 5

A data science team is evaluating an Einstein Discovery model that predicts customer churn. The model shows high overall accuracy (95%) but performs poorly on the minority class (churned customers) with a recall of only 20%. The team wants to improve the model's ability to identify churners. What should they do?

Accepted Answer

Adjust the classification threshold to favor the minority class. The model's high accuracy but low recall on the minority class indicates a class imbalance problem. Adjusting the classification threshold is a direct way to increase recall for churners by making the model more likely to predict churn. This allows the team to trade off some precision for higher recall, aligning with the goal of identifying more churners.

Answer

Increase the model's complexity by adding more features

Answer

Switch to a different algorithm like linear regression

Answer

Remove outliers from the dataset

Question 6

A Salesforce admin is setting up Einstein Lead Scoring and wants to ensure the model generates accurate predictions. Which action should the admin take to improve model accuracy?

Accepted Answer

Ensure that lead records include complete and consistent data in key fields.. Einstein Lead Scoring learns from historical lead data, so the quality and completeness of that data directly impact prediction accuracy. Ensuring key fields are populated and consistent helps the model identify meaningful patterns. Other actions like changing page layouts or manually overriding scores do not enhance the model's learning and may even harm its effectiveness.

Answer

Increase the number of lead fields on the page layout.

Answer

Disable lead scoring for leads with missing values.

Answer

Manually adjust the lead scores after each scoring cycle.

Question 7

A retail company's Einstein Lead Scoring model is trained on two years of Opportunity data. The data science team notices that the model assigns very high scores to leads from one region simply because that region historically generated more revenue, even when the individual lead's engagement is low. Which concept explains why the model is producing this biased pattern?

Accepted Answer

Training data bias. Training data bias occurs when historical data contains patterns that reflect past inequities or structural advantages, and the model learns to replicate them. Because one region historically produced more revenue, the model treats region as a strong signal even for low-engagement leads. Detecting and mitigating this requires examining outcomes across segments and potentially rebalancing data or removing proxy features.

Answer

Overfitting to seasonal demand

Answer

Insufficient feature engineering

Answer

Low model accuracy

Question 8

A sales operations team notices that their Einstein Lead Scoring model is not producing scores for a large portion of newly imported leads. They confirm Einstein Lead Scoring is enabled and that the org has more than the minimum required converted leads. Which condition is most likely preventing scoring for those leads?

Accepted Answer

The leads fall outside the scoring model's active date range or lack sufficient matching fields that the model uses for scoring.. Einstein Lead Scoring generates scores only for leads that fall within the model's trained date range and that contain enough populated fields matching the model's features. Bulk-imported leads that are old or sparsely filled typically show no score. Checking the model's scoring window and data completeness is the correct diagnostic step before assuming a configuration problem.

Answer

The leads have never been converted, and Einstein Lead Scoring only scores leads that already have a converted counterpart in the training set.

Answer

The org has exceeded the maximum number of lead records allowed for Einstein Lead Scoring, so scoring silently stops.

Answer

The leads were imported without a value in the Lead Source field, which the model requires as a mandatory scoring input.

Question 9

A company wants to use Einstein Prediction Builder to predict customer churn. They have a custom object called Subscription__c with historical data. Which requirement must be met to create a prediction for churn?

Accepted Answer

The object must have at least 400 records with a known outcome for the field being predicted.. Einstein Prediction Builder requires at least 400 records with a known outcome for the field being predicted to train a model. This threshold ensures enough data for meaningful predictions. The outcome can be categorical or numeric, and no formula field or Chatter enablement is needed. The correct requirement is the minimum record count.

Answer

The prediction must be based on a numeric field that represents the churn outcome.

Answer

The prediction must use a formula field to calculate churn risk.

Answer

The object must be enabled for Chatter to allow predictions.

Question 10

A Salesforce admin at Northern Trail Outfitters has enabled Einstein Lead Scoring. The sales team is now asking how the feature decides which leads are most likely to convert. Which statement accurately describes what Einstein Lead Scoring does?

Accepted Answer

It uses a predictive model built from the org's historical lead data to assign each lead a score indicating likelihood of conversion.. Einstein Lead Scoring is a predictive feature that trains on the org's historical lead data to estimate how likely each lead is to convert. It surfaces a score along with the top contributing factors so reps can prioritize. The other described mechanisms, such as static rules, external enrichment, or engagement-only signals, describe different tools and do not reflect how this scoring model produces its predictions.

Answer

It applies a fixed set of administrator-defined rules to rank leads based on field values such as industry or annual revenue.

Answer

It queries an external data provider in real time to enrich each lead and then ranks the lead by enrichment completeness.

Answer

It calculates a score from the lead's engagement with marketing emails and website visits only.

Question 11

A service manager at Ursa Major Solar notices that customers write long complaint emails containing both praise and frustration. The manager wants Salesforce to classify the overall tone of each incoming email so that negative messages are routed to senior agents. Which Salesforce AI feature should be used?

Accepted Answer

Einstein Sentiment Analysis. The requirement is to interpret the emotional tone of free-text email content and use it for routing. Einstein Sentiment Analysis is purpose-built for scoring text as positive, negative, or neutral, and those scores can trigger assignment or escalation rules. The other features handle prediction of outcomes, recommendation of actions, or activity logging, none of which classify tone.

Answer

Einstein Prediction Builder

Answer

Einstein Next Best Action

Answer

Einstein Activity Capture

Question 12

A customer service team wants to use Salesforce Einstein to automatically classify incoming cases into categories such as 'Billing', 'Technical', or 'General'. They have a large set of historically labeled cases. Which Einstein capability should they use?

Accepted Answer

Einstein Case Classification. Einstein Case Classification is the correct choice because it is purpose-built to automatically categorize cases based on their content. It uses historical case data to train a model that predicts the appropriate category for new cases. This reduces manual triage and improves routing accuracy. Other Einstein features like Sentiment Analysis, Prediction Builder, or Lead Scoring serve different purposes and cannot perform multi-class text classification for case management.

Answer

Einstein Prediction Builder

Answer

Einstein Sentiment Analysis

Answer

Einstein Lead Scoring

Question 13

A marketing analyst at Northern Trail Outfitters is preparing data to train an Einstein Prediction Builder model that predicts whether a subscriber will make a purchase in the next 60 days. The analyst must identify which data conditions support a reliable model. (Choose two.)

Accepted Answer

Historical records include a sufficient volume of examples spanning multiple past purchase cycles.. A reliable Prediction Builder model requires a well-defined outcome field with both positive and negative examples, plus enough historical records covering representative time periods. A balanced binary outcome gives the model contrast to learn from, and multiple past purchase cycles provide the volume and variation needed for stable scores. Manual review, narrow training sets, or single free-text predictors do not meet these conditions.

Answer

The prediction is trained only on subscribers who already made a purchase in the last 30 days.

Answer

All subscriber records are manually reviewed and corrected by the marketing team before training.

Answer

The model uses a free-text field containing agent notes as the sole predictor variable.

Question 14

A customer support director notices that cases are being routed to the wrong queues because the priority field is often left at the default value by agents. The director wants Salesforce to read the incoming case subject and description and automatically predict the correct priority so routing rules fire immediately. Which Einstein feature should be configured?

Accepted Answer

Einstein Case Classification. Predicting a case field such as priority from the subject and description is the defined purpose of Einstein Case Classification, which uses historical cases to learn patterns and then writes predicted values onto new cases so routing and escalation rules can act. Sentiment, recommendations, and article suggestions produce different outputs and cannot set the priority field.

Answer

Einstein Article Recommendations

Answer

Einstein Next Best Action

Answer

Einstein Sentiment Analysis

Question 15

A service manager at a utility company wants to reduce handle time on case emails. They plan to use Einstein Case Classification to predict the Case Type and Priority on incoming cases. Which statement best describes how Einstein Case Classification produces these predictions?

Accepted Answer

It uses a language model to read the case subject and description and suggests field values learned from previously closed cases.. Einstein Case Classification applies natural language processing to case text and predicts values for configured picklist fields based on patterns learned from historical closed cases. It helps agents triage faster by suggesting Type and Priority. The other options describe manual keyword rules, customer self-selection, or agent routing, none of which reflect how the classification model generates its field predictions from case content.

Answer

It uses a scoring model that ranks agents by their historical resolution speed and routes cases to the fastest agent.

Answer

It requires the customer to select a case type and priority from a self-service portal before the case is created.

Answer

It runs a deterministic keyword mapping that the admin maintains manually in Setup and assigns values when keywords match.

Question 16

A healthcare company is building an AI model to predict patient no-shows for appointments. The data science lead wants to ensure the model generalizes well to new patients rather than memorizing the training examples. Which practice should the team apply during model development?

Accepted Answer

Evaluate the model on a held-out validation set separate from the training data. Generalization is assessed by testing the model on data it has never seen. A held-out validation set provides that unbiased estimate and exposes overfitting, where training performance is high but new-data performance is poor. This practice is fundamental to building models that predict no-shows reliably for future patients rather than memorizing past records.

Answer

Train the model only on the most recent month of appointment data

Answer

Increase the number of features until training accuracy reaches 100 percent

Answer

Remove all patients who missed appointments from the training data

Question 17

A Salesforce admin is reviewing Einstein Opportunity Scoring with the sales operations team. The team asks which statement accurately reflects a limitation or behavior of this feature that they should plan around.

Accepted Answer

The model requires a minimum number of historical opportunities in the org before scores are generated, and new orgs may see no scores initially.. Einstein Opportunity Scoring depends on historical opportunity data to train its model, so orgs with limited closed opportunities may not receive scores until enough data accumulates. This is an important planning consideration for rollout timelines. The score supplements rather than replaces stage management, updates on a regular schedule rather than only on edits, and is governed by normal field-level access rather than an elevated permission.

Answer

Scores are visible only to users with the Modify All Data permission, which prevents most reps from seeing them.

Answer

The score is recalculated only when the opportunity record is edited, so untouched opportunities can display stale scores for long periods.

Answer

The score replaces the opportunity stage, so sales reps no longer need to update the stage field manually.

Question 18

A customer service director at Cosmic Solutions wants an AI feature that recommends the best next action for an agent to take on a case, based on business rules and predictive scores already available in Salesforce. Which capability should the director implement?

Accepted Answer

Einstein Next Best Action. The director wants recommendations for the next action an agent should take, informed by business rules and predictive scores. Einstein Next Best Action is built for that purpose: it evaluates a strategy of rules and offers ranked recommendations in the user interface. Prediction Builder only produces scores, Bots handle conversations, and Activity Capture logs activity, so none of them deliver in-console action guidance.

Answer

Einstein Activity Capture

Answer

Einstein Bots

Answer

Einstein Prediction Builder

Question 19

A Salesforce admin is enabling Einstein Generative AI in their org. To comply with ethical AI guidelines, they enable data masking in the Einstein Trust Layer. Which risk does this primarily mitigate?

Accepted Answer

Sensitive customer data being exposed to the large language model.. Data masking in the Einstein Trust Layer is designed to protect sensitive information by replacing it with placeholders before it reaches the LLM. This prevents exposure of PII and other confidential data, aligning with ethical AI principles of privacy and security. The other options describe risks that masking does not directly mitigate, such as bias, internal access control, or data quality.

Answer

Biased outputs generated by the large language model.

Answer

Unauthorized access to Salesforce records by internal users.

Answer

Inaccurate predictions from the AI model due to poor data quality.

Question 20

A customer support manager wants to reduce handling time by having an AI system automatically route incoming cases to the correct queue based on the text of the case description. The team has thousands of historically labeled cases mapping descriptions to queues. Which type of machine learning best fits this requirement?

Accepted Answer

Supervised classification. Because the team possesses thousands of cases already labeled with their correct queues, the task is a classic supervised classification problem. The model learns the mapping between case text and queue from those labels, then predicts the queue for new cases. This leverages existing data and produces a direct routing decision, unlike clustering or anomaly detection.

Answer

Reinforcement learning

Answer

Unsupervised clustering

Answer

Anomaly detection