Cert Prep: Salesforce AI Associate

User Spoofing

The act of creating a believable online profile using AI-generated images. This allows fake users to interact convincingly with real users, posing challenges for businesses trying to identify networks of bots promoting their own content.

Anthropomorphism

The attribution of human traits, emotions, or intentions to non-human entities, often seen in AI when robots or software are given human-like attributes.

Numeric Predictions

The predictive nature of AI that often assigns values between 0 (indicating an unlikely event) and 1 (indicating a certain event). While it can include percentage values, it's not limited to them; for instance, AI can predict specific numbers, like the amount in dollars.

Tokenization

The process of splitting sentences into individual words or tokens. While straightforward in English due to spaces, languages like Thai or Chinese present challenges, necessitating a deep understanding of vocabulary and morphology.

NLU - Natural Language Understanding

The transformation of unstructured data into structured data. It's essentially the process where machines extract meaning from human input, often seen as a subtask of NLP.

Language Processing

A facet of AI known as Natural Language Processing (NLP) that understands word associations to extract underlying intentions. This capability enables functions like translating a document from English to German or summarizing a lengthy scientific paper.

What role does data play in AI models? A. Data is used for training and testing AI models. B. Data is only used for validating AI models. C. Data is only used for testing AI models.

A. Data is used for training and testing AI models. Training data is used to teach the AI model how to make predictions or decisions while testing data is used to evaluate the model's performance and accuracy.

Classifications

AI's ability to categorize data or patterns, often outperforming human abilities. Each AI classifier is specialized for a distinct task. For instance, an AI designed for detecting phishing emails would not perform well in identifying images of fish.

Yes-and-No Predictions

AI's capability to provide binary answers by analyzing historical data, such as determining if a lead is viable for a business or predicting email engagement.

Bias

An inaccuracy that is systematically incorrect in the same direction and affects the success of the person or algorithm that contains it.

Proxy Values

Attributes in a dataset that correlate with sensitive variables. A strong correlation, such as Account Name being a 90% proxy for postal code, might infer associations between various data points.

Stemming vs. Lemmatization

Both methods aim to reduce words to their base or root form. Stemming might simplify words in a rudimentary manner, which can lead to non-existent stems. Lemmatization, on the other hand, considers the part of speech and provides a more valid base word or lemma.

Generator

In GANs, creates potential data outputs.

ZPD - Zone of Proximal Development

A concept from educational psychology about the difference between what a learner can do with and without help. Not inherently an AI term but can be applied in AI-based adaptive learning systems.

Prediction

A derived value, produced by a model, that represents a possible future outcome.

LLM - Language Learning Models

A general term for models trained on linguistic data.

Sentiment Analysis

A method that evaluates digital text to discern the emotional undertone of a message, categorizing it as positive, negative, or neutral.

A consultant discusses the role of humans in AI-driven CRM processes with a customer. What is one challenge the consultant should mention about human-AI collaboration in decision-making? A. Difficulty in interpreting AI decisions. B. High cost of AI implementation. C. Lack of technical skills in the team.

A. Difficulty in interpreting AI decisions. AI decisions are often based on complex algorithms and large datasets, making them difficult for humans to interpret without sufficient expertise and understanding of AI principles.

What are the three main types of AI capabilities in Salesforce? A. Predictive, Generative, Analytic. B. Predictive, Reactive, Analytic. C. Generative, Descriptive, Analytic.

A. Predictive, Generative, Analytic. Salesforce primarily offers predictive, generative, and analytic AI capabilities.

GPT - Generative Pre-Trained Transformer

An AI tool that has been trained to write like humans do.

Cloud Kicks is implementing AI in its CRM system and is focusing on data management. What is the benefit of using a data management approach in AI implementation? A. Eliminates the need for data governance. B. Reduces the amount of data in the CRM system. C. Emphasizes the importance of data quality.

C. Emphasizes the importance of data quality. Data quality, preparation and cleansing, and data governance are all essential when implementing AI.

Marketing Cloud Einstein

Empowers marketers to better understand their customers. It uncovers consumer insights, suggests optimal engagement channels, personalizes content, streamlines marketing operations, and even auto-generates subject lines and web campaigns.

Comparative Insights in Einstein Discovery

Insights, also derived from the model, elucidate the difference in outcomes by contrasting two specific subgroups. They help in comparing the impact of factors against global averages, with waterfall charts assisting in visualization.

AI - Artificial Intelligence

Intelligence demonstrated by machines, enabling them to perform tasks typically requiring human intelligence.

A consultant designs a new AI model for a financial services company that offers personal loans. Which variable within their proposed model might introduce unintended bias? A. Loan Date. B. Postal Code. C. Payment Due Date.

Postal codes can Introduce bias as they are often correlated with socioeconomic status and race. This Is due to historical practices such as redlining, where certain neighborhoods were marked as hazardous, often denying access to low-cost home lending to minority groups residing In these areas.

Machine Learning Bias

Refers to AI algorithms that inadvertently mirror human prejudices. This phenomenon results in systematically skewed outcomes, usually a consequence of mistaken assumptions during the machine learning process.

Model

Refers to the specific mathematical structure and algorithms used for a particular AI task.

Generative AI

Specialized AI capable of generating media like text or images. These models learn from input data patterns and produce new content resembling the learned characteristics. (AI techniques used to create content or data models.)

Einstein Discovery

Statistical modeling and supervised machine learning in a no-code-required, rapid-iteration environment.

Structured vs. Unstructured Data

Structured data is organized and labeled, like a spreadsheet, making it easily analyzable. In contrast, unstructured data, like a news article or an unlabeled image, lacks a pre-defined structure. The type of data dictates the kind of AI training possible, with unstructured data typically being used for unsupervised learning.

Discriminator (in GAN - In Generative Adversarial Networks)

The discriminator evaluates the authenticity of data.

Salesforce Einstein Discovery Use Cases

The platform addresses specific scenarios like regressions for numeric outcomes (e.g., currency, counts), binary classification for text outcomes with two possible results (e.g., churned or not churned), and multiclass classification for text outcomes with 3 to 10 potential results.

Hallucination

When AI models produce outputs that aren't based on the data they were trained on.

GPT - Generative Pre-trained Transformer

a type of deep learning model specialized in generating text.

Types of AI Capabilities

• Numeric Predictions. • Classifications. • Robotic Navigation. • Language Processing.

Salesforce's Trusted AI Principles

• Responsible. • Accountable. • Transparent. • Empowering. • Inclusive.

Service Cloud Einstein

Centers on enhancing customer service experience. It expedites case resolutions, boosts call deflection, reduces handle times, and offers agents smart suggestions and knowledge recommendations to resolve issues faster.

CRM with AI

Customer Relationship Management systems enhanced with AI features to automate tasks and personalize customer interactions.

Sales Cloud Einstein

Focused on maximizing sales productivity. This tool prioritizes leads and opportunities that are most likely to convert, analyzes sales cycles, automates data capture, and fosters relevant customer outreach based on CRM data.

Generative AI vs. Predictive AI

Generative AI generates new data resembling its training data, while Predictive AI uses historical data to make future predictions.

Grounding

Linking abstract concepts in AI models to concrete instances or data.

Supervised Learning

Machine learning where models learn from labeled data.

Transparency

Making the decision-making process of AI models clear and understandable.

Prompt Engineering

Methods of refining AI model responses through prompt optimization.

Data Leakage - (hindsight bias)

Results from using variables in your prediction model that are known only after the act of interest.

Neural Network

A web connections, guided by weights and biases.

Augmented Intelligence

Combining human intelligence with AI to enhance and scale decision-making capabilities.

Zero Data Retention

Policy of not retaining user data to ensure privacy and security.

What is a key benefit of implementing AI in a CRM system? A. Enhanced customer support. B. Improved platform speed. C. Reduced data governance.

A. Enhanced customer support. Enhanced customer support is a key benefit of implementing AI in CRM.

NLP - Natural Language Processing

A branch of artificial intelligence that combines computer science with linguistics. Its goal is to empower computers with the ability to comprehend, interpret, and reproduce human language in ways meaningful to humans.

Premortem

A pre-project meeting to set measured and realistic expectations and catch the "what went wrong" before it happens.

Algorithm

A predefined set of rules or procedures for solving a problem (turning an input into an output).

Einstein Prediction Service

A public REST API service that facilitates interaction with Einstein Discovery-driven models and predictions. Its capabilities range from fetching predictions on data and suggesting beneficial actions to managing model refresh jobs and bulk scoring tasks.

Commerce Cloud Einstein

A solution for multi-channel brand interaction. It personalizes shopping experiences, offers insights into buying patterns, and optimizes both explicit and implicit search experiences for shoppers. It also supports the automatic generation of smart product descriptions.

Red-Teaming

A strategy of simulating adversarial attacks to identify vulnerabilities in systems.

ML - Machine Learning

A subset of AI that uses vast amounts of data to train models to make predictions. Unlike traditional algorithms which are explicitly programmed, ML models learn from the data they are fed.

Characteristics of an Effective Chatbot

A successful chatbot is transparent, personable, thorough, and continuously evolving. It should clearly identify its nature, embody the brand's tone, provide comprehensive information, and adapt based on feedback.

NBA - Einstein Next Best Action

A system that provides timely and intelligent recommendations using both rules-based and predictive models. These insights are directly integrated within Salesforce for optimal impact.

GAN - Generative Adversarial Network

A type of AI model comprising two parts: a generator and a discriminator.

Generative AI for Business Leaders

A type of AI that's capable of generating new content, ideas, or solutions. This contrasts with predictive or traditional AI, which typically analyzes existing data for patterns. Generative AI can be crucial for innovation and strategy planning.

Reinforcement Learning

A type of machine learning where agents learn by interacting with environments and receiving feedback.

What is a unique and distinguishing feature of deep learning in the context of AI capabilities? A. Deep learning uses neural networks with multiple layers to learn from a large amount of data. B. Deep learning uses historical data to predict future outcomes. C. Deep learning uses algorithms to cleanse and prepare data for AI implementations.

A. Deep learning uses neural networks with multiple layers to learn from a large amount of data. This sets deep learning apart from other types of AI that may not use neural networks or may use them in a different way.

A Salesforce consultant is discussing AI capabilities with a customer who is interested in improving their sales processes. Which type of AI would be most suitable for enhancing sales processes in Salesforce Customer 360? A. Predictive Analytics. B. Computer Vision. C. Natural Language Processing (NLP).

A. Predictive Analytics. This type of AI car enhances sales processes by predicting future outcomes based on historical data.

Cloud Kicks wants to implement Salesforce's AI features. They are concerned about potential ethical and privacy challenges. What should be recommended to minimize potential AI bias? A. Salesforce's Trusted AI Principles. B. Demographic data to identify minority groups. C. AI models that auto-correct biased data.

A. Salesforce's Trusted AI Principles. These principles guide the development and use of AI within Salesforce, ensuring that it is used ethically and responsibly, which includes minimizing potential AI bias.

HITL - Human in the Loop

AI systems incorporating human intervention for decision-making.

NLP - Natural Language Processing

AI that interprets and generates human language.

Robotic Navigation

AI's proficiency in navigating dynamic environments, such as autonomous vehicles adapting to various road conditions, including curves, wind gusts from large trucks, and sudden traffic stops.

Machine Learning

AI's subset where systems learn from data to improve task performance without explicit programming.

Einstein Bots

Allows you to build a smart assistant into your customers' favorite channels like chat, messaging, or voice.

Neural Networks

An AI methodology that simulates the structure and function of the human brain to process information, enabling the machine to learn from and interpret data in a human-like manner.

NER - Named Entity Recognition

An NLP technique that uses algorithms to identify and classify named entities such as people, dates, places, and organizations within text. This aids in tasks like question answering and information extraction.

Syntactic Parsing

An analysis technique that interprets the words in a sentence based on grammar and arranges them to illustrate relationships among the words. Sub-components include segmentation, tokenization, stemming, lemmatization, part of speech tagging, and named entity recognition.

Speech Tagging

An essential function in NLP where each word in a sentence is labeled based on its grammatical role, such as noun, verb, adjective, etc. This helps in understanding the syntax and meaning of a sentence.

Einstein Prediction Builder

An intuitive tool that allows for custom predictions on non-encrypted Salesforce data across various business sectors. It emphasizes user-friendly functionality without the need for coding.

Deep Learning and Data

Analogous to the human mind's ability to establish connections and derive insights, deep learning employs artificial neural networks across large databases. Using algorithms, it sifts through information, derives conclusions, and refines performance.

Einstein Agent Productivity

Armed with predictive intelligence, this feature offers real-time suggestions to agents, equipping them to deliver exemplary customer service.

Which AI type plays a crucial role in Salesforce's predictive text and speech recognition capabilities, enabling the platform to understand and respond to user commands accurately? A. Computer Vision. B. Natural Language Processing (NLP). C. Predictive Analytics.

B. Natural Language Processing (NLP). NLP enables computers to understand, interpret, and generate human language in a meaningful way.

Prompt Defense

Techniques ensuring AI models respond safely to given prompts.

A Salesforce consultant is considering the data sets to use for training AI models for a project on the Customer 360 platform. What should be considered when selecting the data sets for the AI models? A. Duplication, accuracy, consistency, storage location, and usage of the data sets. B. Age, completeness, consistency, theme, duplication, and usage of the data sets. C. Age, completeness, accuracy, consistency, duplication, and usage of the data sets.

C. Age, completeness, accuracy, consistency, duplication, and usage of the data sets. These are the key elements/components of data quality that are crucial when selecting data sets for AI models.

Cloud Kicks wants to implement AI features within its CRM system. They have expressed concerns about the quality of their existing data. What advice should be given to them regarding the importance of data quality for AI implementations? A. Assessing data quality is only necessary for large datasets. B. AI systems can handle any data inaccuracies. C. Assessing and improving data quality is crucial for accurate AI predictions and insights.

C. Assessing and improving data quality is crucial for accurate AI predictions and insights. Assessing and improving data quality is crucial for accurate AI predictions and insights.

Which Salesforce AI application is recommended to enhance sales processes? A. Einstein Prediction Builder. B. Einstein Voice. C. Einstein Lead Scoring.

C. Einstein Lead Scoring. Einstein Lead Scoring is specifically designed to enhance sales processes by scoring leads based on their likelihood to convert, allowing sales teams to prioritize their efforts effectively.

Which feature of Marketing Cloud Einstein uses AI to predict consumer engagement with email and MobilePush messaging? A. Content Selection. B. Email Recommendations. C. Engagement Scoring.

C. Engagement Scoring. Customer data and machine learning are used to assign scores for every contact 's likelihood to engage with emails and interact with push notifications.

Cloud Kicks is planning to automate its customer service chat using natural language processing. According to Salesforce's Trusted AI principles, how should this be disclosed to the customer? A. They do not need to be informed they are chatting with AI. B. Inform the customer that they are chatting with AI when they request a live agent. C. Inform them at the beginning of the interaction that they are chatting with AI.

C. Inform them at the beginning of the interaction that they are chatting with AI. This allows customers to understand the context of their interaction and sets appropriate expectations.

Elements of Natural Language in English

Components that make up the language including vocabulary, grammar, syntax, semantics, pragmatics, discourse, dialogue, phonetics, phonology, and morphology.

Artificial Neural Network

Computational models inspired by the human brain's structure, consisting of interconnected nodes (neurons) used for pattern recognition and decision-making.

Parameters

Configurable variables in a model that are learned from the training data to make predictions.

Which data quality dimension refers to the frequency and timeliness of data updates? A. Data Source. B. Data Freshness. C. Data Leakage.

Data freshness refers to how up to date or current the data is, which includes the frequency and timeliness of data updates.

Descriptive Insights in Einstein Discovery

Derived from historical data via descriptive analytics and statistical analysis, these insights depict what has transpired in your data.

Disparate Impact

Discriminatory practices toward a particular demographic.

Einstein Discovery Insights

Einstein's Discovery explores patterns, relationships, and correlations in historical data. Using machine learning and AI, it predicts future outcomes, guiding business users in prioritizing their tasks and making informed decisions.

Validation

Evaluating the performance of an AI model on a separate dataset not used during training.

Data Insights

Findings in your data bases on thorough analysis generated by Einstein Discovery.

Ethical AI Maturity Model

Framework for ensuring AI systems is developed and deployed ethically and responsibly.

Toxicity

Harmful outputs or behaviors in AI models.

Salesforce Data Cloud

Harmonizes and stores customer data at massive scale, and transforms it into a single, dynamic, source of truth

Diagnostic Insights in Einstein Discovery

Insights, originating from the model, delve deep to explain why certain events occurred. They identify which variables most significantly influence the analyzed business outcome.

Predictive AI

Studies historical data, identifies patterns, and makes predictions about the future that can better inform business decisions. Predictive AI's value is shown in the ways it can detect data flow anomalies and extrapolate how they will play out in the future in terms of results or behavior; enhance business decisions by identifying a customer's purchasing propensity as well as upsell potential; and improve business outcomes

Einstein Discovery-Powered Solutions

These solutions leverage the power of Einstein Discovery for various business scenarios. Examples include predicting numeric outcomes, classifying binary text outcomes, and multi-class text classifications.

LLMs - Large Language Models

Trained on vast amounts of textual data from the internet, these models execute advanced language tasks such as summarizing, translating, and error correction, among others.

Sensitive Variables in Einstein Discovery

Variables related to legally protected classes, such as age, race, and gender, have use restrictions in regions like the US and Canada. Discrimination against these classes in sectors like employment, healthcare, and lending is prohibited. Proxy values can correlate with these sensitive variables, potentially leading to biased interpretations if not handled correctly.