AI Associate Practice Exam 55 — 20 Questions

Question 1

A Salesforce admin is preparing a dataset for an AI model that predicts customer satisfaction. The dataset includes a field 'Region' with values like 'North', 'South', 'East', and 'West'. The admin wants to ensure the model can use this field effectively. What should the admin do?

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

Accepted Answer

One-hot encode the 'Region' field into separate binary columns for each region.. One-hot encoding is the standard method for handling nominal categorical variables in machine learning. It creates binary columns for each category, allowing the model to treat each region independently without imposing an order. This is crucial for fields like 'Region' where no inherent ranking exists. Proper encoding improves model accuracy and avoids misleading the algorithm.

Answer

Convert 'Region' to a numeric field by mapping each region to its average satisfaction score.

Answer

Label encode the 'Region' field by assigning numeric values (e.g., North=1, South=2).

Answer

Remove the 'Region' field because it is not a numeric feature.

Question 2

A data engineer at a retail company is preparing a dataset in Salesforce Data Cloud to train a propensity model. The dataset contains a field named 'Preferred_Contact_Time' with values such as 'Morning', 'Afternoon', 'Evening', and 'Weekend'. The model requires numerical input, but the field is stored as text. The engineer wants to transform this field into a format suitable for the model. Which approach should the engineer take?

Accepted Answer

Use one-hot encoding to create binary columns for each category.. Categorical text data must be encoded into numerical representations before being used in most AI models. One-hot encoding creates separate binary columns for each category, avoiding any implied order and allowing the model to treat each category independently. This is the recommended method for nominal data like preferred contact time, ensuring the model can learn without bias from arbitrary numeric assignments.

Answer

Apply label encoding to assign integer values (e.g., Morning=1, Afternoon=2).

Answer

Replace the text values with their frequency counts in the dataset.

Answer

Convert the text to lowercase and use it directly as a numerical feature.

Question 3

A data scientist is preparing a dataset for an Einstein Discovery model to predict customer churn. The dataset includes a 'CustomerSince' date field, a 'TotalSpend' numeric field, and a 'SupportTickets' numeric field. The data scientist wants to create new features that capture customer tenure and engagement. Which two feature engineering techniques should be applied? (Choose two.)

Accepted Answer

Compute the ratio of TotalSpend to SupportTickets to create a 'SpendPerTicket' feature.. Creating a tenure feature from CustomerSince and a spend-per-ticket ratio from TotalSpend and SupportTickets are both effective feature engineering techniques. They transform raw data into meaningful predictors that can help the model identify churn patterns. These features capture customer loyalty and engagement, which are key drivers of churn.

Answer

Bin the TotalSpend field into quartiles and treat as categorical.

Answer

Create a polynomial feature by squaring SupportTickets.

Answer

Apply one-hot encoding to the CustomerSince date field.

Question 4

A data scientist is preparing a dataset for an Einstein Discovery model that predicts customer satisfaction scores. The dataset includes a field 'Region' with values like 'North', 'South', 'East', 'West', and 'Unknown'. The data scientist notices that 'Unknown' appears in 30% of records. What is the most appropriate action regarding this field?

Accepted Answer

Keep 'Unknown' as a separate category and let the model determine its impact.. Keeping 'Unknown' as a separate category allows the model to learn any predictive signal associated with missing region data. With 30% missing, imputation or removal could introduce bias and lose information. Einstein Discovery can handle categorical missing values effectively, making this the most appropriate action.

Answer

Remove all records with 'Unknown' region to ensure data quality.

Answer

Create a new binary field indicating whether the region is known or unknown, and drop the original 'Region' field.

Answer

Replace 'Unknown' with the most frequent region to reduce missing data.

Question 5

A data scientist is building a predictive model in Salesforce Data Cloud to forecast next quarter's sales. The dataset includes historical sales, marketing spend, and economic indicators. The data scientist notices that the 'Marketing_Spend' feature has a high correlation with the target 'Sales' but also with another feature 'Economic_Indicator'. The model's performance is unstable, with high variance in predictions. Which issue is most likely causing this instability?

Accepted Answer

Multicollinearity between 'Marketing_Spend' and 'Economic_Indicator'.. Multicollinearity arises when independent variables are highly correlated, making it hard for the model to determine the individual impact of each feature. This leads to unstable coefficient estimates and high variance in predictions. In this case, the correlation between marketing spend and economic indicator is likely causing the instability. Techniques like removing one of the correlated features, using principal component analysis, or applying regularization can mitigate the issue.

Answer

Overfitting due to too many features relative to observations.

Answer

Data leakage from including future economic indicators.

Answer

Underfitting because the model is too simple to capture the relationship.

Question 6

A Salesforce admin is preparing a dataset for an AI model that predicts customer satisfaction. The dataset includes a field 'Region' with values like 'North', 'South', 'East', and 'West'. The admin wants to ensure the model can use this field effectively. What should the admin do?

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

Accepted Answer

Standardize the date format to a consistent ISO 8601 format (YYYY-MM-DD) using a data preparation tool or formula field.. Inconsistent date formats prevent Einstein Discovery from recognizing the field as a valid date type, leading to errors or poor model performance. Standardizing to a uniform format like ISO 8601 ensures the model can correctly parse and use the temporal information. Other options either skip this critical step or remove valuable data.

Answer

Convert the date field to a numeric value by calculating the number of days since a fixed reference date.

Answer

Create a new text field that concatenates the date with a customer ID to create a unique identifier.

Answer

Remove the 'LastPurchaseDate' field entirely because date fields are not supported by Einstein Discovery.

Question 7

A Salesforce admin is preparing a dataset for an AI model that predicts customer satisfaction. The dataset includes a field 'Region' with values like 'North', 'South', 'East', and 'West'. The admin wants to ensure the model can use this field effectively. What should the admin do?

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

Accepted Answer

Group rare categories into an 'Other' category to reduce cardinality.. High cardinality in categorical fields can cause overfitting and reduce model interpretability. Grouping rare categories into an 'Other' bucket reduces the number of unique values while retaining the most informative ones. This preparation step improves model performance and is a best practice for fields like 'LeadSource' with many free-text entries.

Answer

Convert the 'LeadSource' field to a numeric field using label encoding.

Answer

Remove the 'LeadSource' field entirely because it has too many unique values.

Answer

Leave the field as is, because Einstein Discovery automatically handles high-cardinality fields.

Question 8

A data engineer at a retail company is building a propensity-to-buy model in Einstein Studio using Data Cloud. The training dataset includes a 'LastPurchaseDate' field. The engineer notices that many records have purchase dates from 2018, while the model is intended to predict purchases for the upcoming holiday season in 2024. Which data quality dimension is most directly compromised?

Accepted Answer

Timeliness. Timeliness is the data quality dimension that ensures data is current enough for its intended use. In this predictive modeling scenario, using purchase dates from 2018 to forecast 2024 behavior introduces staleness. The model may learn outdated patterns that no longer apply, reducing predictive accuracy. Addressing timeliness by filtering to recent transactions or creating recency features helps align the training data with the prediction context.

Answer

Consistency

Answer

Accuracy

Answer

Completeness

Question 9

A data scientist is building an Einstein Discovery model to predict opportunity win rates. The dataset includes a 'Probability' field that is manually entered by sales reps and often left blank or set to default values. The data scientist wants to improve model accuracy. What is the most appropriate action regarding the 'Probability' field?

Accepted Answer

Exclude the 'Probability' field from the model because it is manually entered and unreliable.. The 'Probability' field is manually entered, often blank, and set to defaults, making it unreliable and potentially biased. Excluding it prevents the model from learning from noisy data and encourages use of more objective predictors. While imputation or replacement could be considered, they are not the most appropriate first step given the field's issues.

Answer

Impute missing values with the mean probability and keep the field as a predictor.

Answer

Keep the field but create a binary indicator for whether it was manually entered or defaulted.

Answer

Replace the 'Probability' field with a calculated field that uses historical win rates by stage.

Question 10

A Salesforce admin is preparing a dataset for an AI model that predicts customer satisfaction. The dataset includes a field 'Region' with values like 'North', 'South', 'East', and 'West'. The admin wants to ensure the model can use this field effectively. What should the admin do?

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

Accepted Answer

Use Einstein Discovery's built-in missing value handling by leaving the field as is, since the platform automatically imputes or excludes missing values during training.. Einstein Discovery has built-in mechanisms for handling missing values, so the admin can leave the CSAT field as is and let the platform impute or exclude missing entries during training. Deleting records, using artificial placeholders, or discarding the field all introduce bias or lose predictive signal. The platform's native handling is the intended approach for this common data quality issue.

Answer

Replace missing CSAT values with a placeholder like -1 and treat the field as numeric.

Answer

Create a formula field that converts CSAT to a binary 'responded' indicator and drop the original CSAT field.

Answer

Delete all records with missing CSAT values so the model trains only on complete data.

Question 11

A data engineer is building a Data Cloud data model for an Einstein Discovery model that predicts whether a customer will renew a subscription. The source system sends the renewal date in the format MM/DD/YYYY as a text string. The engineer notices that Einstein Discovery is treating this field as a categorical variable instead of a date. What should the engineer do to correct this?

Accepted Answer

In Data Cloud, edit the data model object field mapping and set the field's data type to Date, then refresh the data stream.. Changing the field's data type to Date on the Data Cloud data model object is the correct fix because it controls how Einstein Discovery interprets the variable. Refreshing the stream after the mapping change ensures the model uses the corrected type. Source-side conversions and calculated insights are indirect workarounds that do not resolve the semantic type at the data model layer.

Answer

Use a Data Cloud calculated insight to cast the text field to a date and then create a new model using the insight output.

Answer

Add a filter in Einstein Discovery to exclude the renewal date field from the model because text fields cannot be used in predictions.

Answer

Create a formula field in the source system that converts the text to a date, then re-ingest the data into Data Cloud.

Question 12

A telecommunications company is preparing data for an Einstein Studio churn prediction model. The data engineer notices that the 'CustomerTenure' field has a skewed distribution: most customers have tenure under 2 years, but a few have tenure over 20 years. The engineer wants to improve model performance by addressing this skew. Which data preparation technique should the engineer apply?

Accepted Answer

Apply a log transformation to the 'CustomerTenure' field.. Log transformation is a standard technique to reduce right skewness in continuous features like customer tenure. By compressing the upper tail, it makes the distribution more normal, which helps many machine learning algorithms perform better. This preserves all data points while mitigating the disproportionate influence of extreme values. The engineer should apply this transformation in Data Cloud or Einstein Studio's data preparation steps before training the churn model.

Answer

Remove all customers with tenure over 20 years as outliers.

Answer

Convert 'CustomerTenure' into a categorical field with bins.

Answer

Normalize the 'CustomerTenure' field to a 0-1 range.

Question 13

A financial services company uses Data Cloud to create a unified customer profile for an AI model that predicts loan default risk. The data engineer notices that the 'Income' field from the CRM system and the 'AnnualIncome' field from an external data source frequently disagree for the same customer. Which action should the engineer take to best address this data quality issue for the AI model?

Accepted Answer

Use Data Cloud's reconciliation rules to define a priority order for the income fields.. Data Cloud's reconciliation rules enable data engineers to define which source system's field values take precedence when creating a unified profile. By setting a priority order for income fields, the engineer resolves conflicts consistently and systematically. This ensures the AI model receives a single, trusted income value, which is critical for accurate loan default risk predictions. Relying on manual updates or averaging is not scalable or reliable.

Answer

Exclude both income fields from the model to avoid inconsistency.

Answer

Create a formula field that averages the two income values.

Answer

Manually update the CRM records to match the external source values.

Question 14

A Salesforce data architect is preparing data for an Einstein Discovery model that predicts opportunity win probability. The source data includes opportunity amounts, close dates, stages, and account industry. The architect wants to ensure the model has high-quality training data. Which two actions should the architect take? (Choose two.)

Accepted Answer

Ensure that the win/loss label field is populated for all historical opportunities used in training.. High-quality training data requires a fully populated label field so the model can learn from all historical outcomes, and consistent predictor data types with managed outliers so the model interprets features correctly. Including high-cardinality identifiers like owner name or opportunity ID leads to overfitting, and arbitrary date cutoffs are not a data quality best practice.

Answer

Include the opportunity owner's name as a predictor to capture individual sales performance.

Answer

Remove all opportunities that were closed more than five years ago to keep the model current.

Answer

Add the opportunity's unique ID as a predictor to help the model differentiate records.

Question 15

A data scientist is building an Einstein Discovery model to predict customer lifetime value (CLV). The dataset includes a field 'TotalPurchaseAmount' that is highly skewed, with a few customers having extremely high values. The scientist wants to reduce the impact of outliers on the model. Which data transformation technique should they apply?

Accepted Answer

Logarithmic transformation. Logarithmic transformation is specifically designed to handle skewed data by compressing the scale of large values. It reduces the impact of outliers and makes the distribution more symmetric, which can enhance model performance. Other techniques like normalization or standardization do not mitigate outlier effects and can even exacerbate them.

Answer

One-hot encoding

Answer

Normalization (min-max scaling)

Answer

Standardization (z-score normalization)

Question 16

A manufacturer wants to improve demand forecasting by enriching its CRM orders with external demographic data. The external data is available via a SOAP API. How should the data architect implement this?

Accepted Answer

Configure a Data Action to call the API on a schedule or trigger. Option A is correct because a Data Action is the Pega mechanism designed to invoke external services such as a SOAP API, either on a schedule or triggered by an event, to enrich CRM order data with demographic information. Data Actions can map the SOAP response fields directly into the case or data model for use in demand forecasting. Option B is incorrect because a Data Transform only manipulates data already available in the application and cannot call an external API. Option C is incorrect because a Data Stream is used for batch or real-time ingestion of data into a repository, not for on-demand SOAP service calls. Option D is incorrect because a Calculated Insight computes derived values from existing data and does not invoke external APIs.

Answer

Use a Data Transform to pull the external data

Answer

Set up a Data Stream to continuously ingest the external API

Answer

Create a Calculated Insight to reference the API

Question 17

An organization is implementing Einstein AI for sales forecasting. They have multiple custom objects and complex approval processes. Which design consideration is most critical for ensuring accurate AI predictions?

Accepted Answer

Implementing model explainability to verify that predictions align with business logic.. The correct option is D: implementing model explainability to verify that predictions align with business logic. In an Einstein AI sales-forecasting scenario with multiple custom objects and complex approval processes, prediction accuracy depends less on raw volume or external infrastructure than on being able to inspect which fields and patterns drive each prediction, so stakeholders can confirm the model reflects real business rules and correct data issues. Explainability also supports trust and governance, which is essential when forecasts influence approvals and decisions. Option A is wrong because record-count thresholds alone do not guarantee accuracy and Einstein models have their own data requirements. Option B is unnecessary because Einstein can work with Salesforce data directly without a separate warehouse. Option C addresses user adoption, not prediction accuracy.

Answer

Ensuring the data set includes at least 100,000 records per prediction field.

Answer

Creating a separate data warehouse to aggregate data from all objects.

Answer

Providing extensive user training on AI concepts.

Question 18

Which Einstein feature would allow a company to automatically generate personalized email content for marketing campaigns?

Accepted Answer

Einstein GPT. Einstein GPT is the correct option because it is Salesforce's generative AI capability that can automatically draft personalized email content, such as subject lines and body copy, for marketing campaigns. It uses large language models to generate text based on prompts and customer data, which directly matches the requirement for automatically generating personalized email content. Einstein Segmentation, by contrast, is used to divide audiences into groups rather than create content. Einstein Prediction Builder is designed to build predictive models and score records, not generate email copy. Einstein Analytics (now CRM Analytics) focuses on data visualization and reporting, so it also does not generate personalized email content.

Answer

Einstein Segmentation

Answer

Einstein Prediction Builder

Answer

Einstein Analytics

Question 19

A financial institution must ensure that customer data used for AI models does not expose personally identifiable information (PII) to unauthorized users. Which Data Cloud feature should be applied to the data model?

Accepted Answer

Apply data masking and field-level security on sensitive fields. Data Cloud supports data masking and field-level security, which allow sensitive PII fields to be obfuscated or restricted while still permitting the data to be used in AI models by authorized users. This directly addresses the requirement to prevent unauthorized exposure of PII without destroying the underlying data model. Option A is too destructive and would remove data needed for AI processing, while Option B only aggregates data and does not protect field-level PII exposure. Option D is insufficient because object-level permissions do not restrict access to individual sensitive fields within an object.

Answer

Delete PII fields from the data model

Answer

Use Calculated Insights to aggregate sensitive data only

Answer

Rely on user permissions to restrict access to the entire object

Question 20

A hospital uses an AI triage system to prioritize patients in the emergency department. The AI was trained on historical patient data and assigns priority scores based on vital signs and symptoms. Recently, a study finds that the system consistently assigns lower priority to elderly patients compared to younger patients with similar clinical presentations. The hospital's ethics committee is concerned about age discrimination. The current model achieves high accuracy in predicting outcomes, and doctors have come to rely on it for efficiency. What should the hospital do to address the ethical concern while maintaining clinical effectiveness?

Accepted Answer

Retrain the model with a modified objective that penalizes age-based disparities.. Retraining the model with a fairness-aware objective that penalizes age-based disparities directly targets the observed bias while preserving the model's predictive utility and clinical workflow. This approach allows the hospital to mitigate discriminatory prioritization without discarding the system's demonstrated accuracy and efficiency. Option A is overly disruptive and unnecessary, as the existing model can be adjusted rather than rebuilt from scratch. Option C ignores the ethical and potentially legal problem of age discrimination, and Option D is insufficient because removing age as an explicit feature does not eliminate bias from correlated proxies such as vital signs, symptoms, or historical treatment patterns.

Answer

Replace the AI triage system with a completely new model built from scratch.

Answer

Continue using the current model since it has high accuracy and efficiency.

Answer

Remove age as an input feature from the model.