Salesforce Einstein AI Features practice questions

Keyword tracking

Why wrong: Keyword tracking is provided by Einstein Conversation Insights, so it is not the correct answer.

Talk time metrics

Why wrong: Talk time metrics are provided, so it is not the correct answer.

Sentiment analysis

Why wrong: Sentiment analysis is indeed provided by Einstein Conversation Insights, so it is not the metric that is NOT provided.

Einstein Opportunity Scoring

Why wrong: This scores opportunities, not leads.

Einstein Prediction Builder

Why wrong: This requires building a custom prediction model, not a pre-built solution for lead scoring.

Einstein Next Best Action

Why wrong: This recommends actions, not scores leads.

Einstein Copilot

Why wrong: Copilot is a conversational assistant, not a prompt management tool.

Agent Builder

Why wrong: Agent Builder is for creating autonomous agents, not prompt templates.

Einstein Studio

Why wrong: Einstein Studio is for building custom AI models, not prompt templates.

Einstein Lead Scoring

Why wrong: Lead Scoring is a pre-built model, not custom. It cannot use arbitrary fields like a custom 'Converted' field.

Einstein Discovery

Why wrong: Discovery provides statistical analysis and stories, not predictive models for lead conversion.

Einstein Opportunity Scoring

Why wrong: Opportunity Scoring is for opportunities, not leads.

The model is overfitting to the training data.

Why wrong: Overfitting would cause high accuracy on training but poor on new data, not necessarily a bias to one class.

The model has not been retrained in over 90 days.

Why wrong: While retraining is important, the symptom described is classic class imbalance, not staleness.

The field mapping for priority is incorrect in the classification configuration.

Why wrong: Incorrect mapping would likely cause errors or blank values, not systematic bias to one class.

Einstein Prediction Builder

Why wrong: Custom binary predictions, not pre-built opportunity analysis.

Einstein Forecasting

Why wrong: Forecasting predicts revenue, not win factors or suggestions.

Einstein Lead Scoring

Why wrong: For leads, not opportunities.

Train a custom model from scratch on the policy documents each month

Why wrong: Training from scratch requires massive compute resources and weeks of time — entirely disproportionate for monthly document updates.

Use a larger foundation model with a longer context window and paste all documents into each prompt

Why wrong: Pasting all documents into every prompt is expensive, hits context limits for large document sets, and does not scale as the document library grows.

Fine-tune a base LLM on the policy documents monthly

Why wrong: Fine-tuning is expensive and time-consuming; monthly cycles are impractical and fine-tuned knowledge becomes stale immediately after cutoff.

The prediction field contains data leakage (e.g., a future value)

Why wrong: Leakage would inflate scores in training and likely production, not cause low confidence on new cases.

The admin selected too few features, causing underfitting

Why wrong: Underfitting yields low training score, not high.

The training data is stale and no longer reflects current case patterns

Why wrong: Stale data would cause poor performance overall, but the training score would also be low.

Einstein Activity Capture

Why wrong: Activity Capture logs emails and events automatically but does not analyze importance.

Einstein Lead Scoring

Why wrong: Lead Scoring predicts conversion likelihood, not email importance.

Einstein Opportunity Scoring

Why wrong: Opportunity Scoring predicts deal win probability, not email prioritization.

The model includes a numeric custom field as a feature that skews the output

Why wrong: Features influence prediction but don't change the output scale.

The prediction field was incorrectly mapped to a currency field

Why wrong: Currency mapping doesn't cause a 0-1000 score; the model creates a numeric field.

The model was trained on too few records, causing overfitting

Why wrong: Overfitting affects accuracy, not the score scale.

Einstein Vision

Why wrong: Vision deals with image classification, not binary predictions from CRM data.

Einstein Next Best Action

Why wrong: Next Best Action recommends actions, not predictions.

Einstein Discovery

Why wrong: Discovery provides analysis and prescriptions, not a binary prediction field.

NLP training

Why wrong: NLP training is important for bot understanding but is a broader prerequisite, not a capability specifically essential for the hybrid escalation model.

Bot analytics

Why wrong: Bot analytics provides performance insights but is not a core capability required for the bot to handle queries and escalate.

Einstein Conversation Insights

Why wrong: Einstein Conversation Insights is a post-interaction analytics tool, not a real-time capability needed for hybrid bot operation.

Einstein Case Classification

Why wrong: Case Classification auto-classifies cases into fields like Type, not custom binary predictions.

Einstein Next Best Action

Why wrong: Next Best Action recommends actions based on strategies, not custom predictions.

Einstein Discovery

Why wrong: Discovery provides statistical analysis and insights but does not create custom prediction models.

Einstein Discovery

Why wrong: Discovery provides analysis and prescriptions, not custom binary prediction models.

Einstein Next Best Action

Why wrong: Next Best Action recommends actions, not predictions.

Einstein Case Classification

Why wrong: Case Classification auto-classifies case fields, but not custom escalations.

Einstein Opportunity Scoring

Why wrong: Opportunity Scoring is for existing opportunities, not leads.

Einstein Activity Capture

Why wrong: Activity Capture logs emails/events; it does not score leads.

Einstein Prediction Builder

Why wrong: Prediction Builder allows custom predictions but is overkill for standard lead scoring.

Prediction Category field, containing the predicted category label

Why wrong: The binary outcome (e.g., Yes/No) is derived from the score, but the score field is the primary new field.

Prediction Confidence field, containing a confidence level of the model

Why wrong: Confidence is not stored as a field; the prediction score itself conveys likelihood.

Prediction Explanation field, containing a text explanation of the prediction

Why wrong: Explanation is available as a separate field or via hover, but the main new field is the score.

Ask the sales reps to manually copy pricing into each prompt.

Why wrong: Manual entry is error-prone and defeats automation purposes.

Use Einstein Recommendation Builder to suggest pricing.

Why wrong: Recommendation Builder is for product recommendations, not for injecting external data into prompts.

Use Einstein Discovery to generate pricing insights.

Why wrong: Einstein Discovery analyzes historical data, it cannot inject real-time external data into email drafts.

Set the prediction window to the next 30 days.

Why wrong: Prediction window is for forecasting future events, not for classifying past data.

Select at least 20 features from the case object.

Why wrong: There is no minimum feature count; feature selection should be relevant, not numerous.

Ensure the data set contains at least 500 records.

Why wrong: While more data helps, the immediate requirement is both classes present. Even with many records, if all are escalated, the model fails.

Einstein Prediction Builder

Why wrong: Prediction Builder works with structured Salesforce data, not images.

Einstein Bots

Why wrong: Bots handle conversational interactions, not image analysis.

Einstein GPT

Why wrong: Einstein GPT generates text, not image classification.

Custom Instructions

Why wrong: Instructions are optional guidelines; not mandatory for agent configuration.

AI Model Selection

Why wrong: The underlying AI model is pre-selected by Salesforce; admins do not choose models.