Salesforce Certified AI Associate Credential

5 Core Functions of Salesforce Einstein

1. Discover insights about customers from data in and out of SF. 2. Predict outcomes tailored to your customer. 3. Recommend optimal actions. 4. Automate routine tasks. 5. Generate tailored content (emails, articles, code).

AI Robotic Navigation

AI that can adapt to changing environmental conditions. Examples include autonomous self-driving cars and vacuum robots.

Risks of using Generative AI

Accuracy: GAI creates new content, but it may not always reflect reality. Bias & Toxicity: AI can replicate and even amplify human biases at scale. Privacy & Safety: Potential misuse for phishing, voice replication, and data leaks. Disruption: Economic shifts, job changes, sustainability concerns due to computing demands. Guidelines & Actions: Product Design: Build with user safety and reliability in mind. Mindful Friction: Educate users on potential biases or concerns. Red Teaming: Test products for vulnerabilities. Acceptable Use Policy: Transparent guidelines on appropriate AI use.

Data Analytics Types

Descriptive Analysis: Reports what happened, using metrics like KPIs. E.g., Time for payment processing in ecommerce. Diagnostic Analysis: Understands why something happened by correlating KPIs. Predictive Analysis: Predicts future events based on past data using techniques like neural networks, regression. Prescriptive Analysis: Suggests actions based on past events using machine learning to analyze large datasets. Data analytics helps improve data-driven decision making

Accountability in Design

Design practices should incorporate both business and social accountability. Business Accountability: Metrics linking internal work to revenue & optimization. Social Accountability: Impact of decisions on society; enhancing positive & mitigating negatives. Prioritizing social value from a design's inception ensures cost-effective inclusion and positive societal impact.

Relationship Design - Design Beyond the User

Designing beyond the user involves considering both the social context (the structures people exist within) and social dynamics (how people interact). While Human-Centered Design focuses on individual needs, Relationship Design emphasizes broader connections and contexts. Ignoring these broader considerations can lead to ineffective products or even harm communities. Important term - Externalities: Unintended side effects impacting those not involved in the product or service creation.

Relationship Scaling Best Practices

No One Solution: Tailor technology solutions to unique user and community needs. Engage with Customers: Understand user interactions, offer value, and stay updated on market needs. Consider Sentiment: Technology can't fully grasp human emotion nuances; prioritize understanding and adapting to diverse human sentiments.

How can AI assist Retail & Commerce

Personalization: -Offers product recommendations based on historic data. -Tailors site content based on past customer purchases. -Reorganizes search results to highlight relevant items. Insights for Merchandisers: -Identifies commonly bundled products for promotions. -Understands customer-site interactions and top referral searches. -Enables crafting of shopping experiences to match customer preferences.

Suggested GDPR Controls and Processes

Privacy notices: Privacy notices must be provided wherever personal data is collected, including through the use of website cookies and tags. Usage limitations: Administrative or technological controls can be used to limit the organization's use of data to the purposes for which it collected the data. Security: Administrative, physical, and technological security measures are necessary to prevent unauthorized access, use, modification, disclosure, or deletion of personal data. Data subject rights: Mechanisms and procedures are needed to manage data subject consent preferences and respond to complaints and requests for access, rectification, restriction, portability, and deletion. Vendor management: Organizations must have contracts with affiliates, vendors, and other third parties that collect or receive personal data, including standard contractual clauses or other mechanisms to legalize data transfers outside the EU. Incident response: Processes must be created to detect and respond to security breaches, including remediating the breach and notifying all necessary parties. Training: Employee and vendor training must be delivered to raise awareness regarding privacy policies, processes, and requirements, as well as to report concerns and suspicious data activity. Assessments: Data protection impact assessments must be conducted for each high risk data processing activity.

How does Salesforce Support the GDPR

SF Has robust security measures and certifications in place. For GDPR compliance, Salesforce highlights the importance of company-wide awareness, thorough organizational assessment, establishing strict controls and processes, and documenting all compliance efforts. The protection of personal data is emphasized as a fundamental right.

Salesforce Data Cloud

Salesforce Data Cloud is a real-time platform that centralizes customer data from diverse sources, creating a unified, dynamic customer graph. It seamlessly integrates this data into the Customer 360 application suite, enhancing automation, intelligence, and engagement across industries. The result is real-time, personalized experiences for customers

California Consumer Privacy Act (CCPA)

A California law passed to protect resident's (including employees, customers, vendors, and contractors) privacy. Any for-profit company that receives the Personal Information of a California resident (a Consumer) and meets any of the following must comply with the CCPA. -Has an annual revenue of $25 million dollars -Handles the data of 50,000 California consumers or devices -Derives 50% or more of its revenue from Selling Personal Information

AI Language Processing

Also known as Natural Language Processing (NLP). Generative AI built to interpret everyday language and act on it in some meaningful way. Natural language processing relies on an understanding of how words are used together then letting AI to extract the intention behind the words. Examples include ChatGPT or Google Bard.

Natural Language Processing (NLP)

Field of AI combining computer science & linguistics to let computers understand, interpret, and generate human language meaningfully. Examples: Email suggestions, virtual assistants (Siri, Alexa), chatbots, spam detection, translations, news preferences.

AI Numeric Predictions

Forecasting or estimation of numerical values based on historical data, patterns, or other inputs. Examples include AI created pricing models for airfare, hotel rooms or insurance based on supply and demand.

Generative AI vs. Predictive AI

Generative AI creates unique work based on analyzed patterns, while Predictive AI uses patterns to make forecasts. Both use machine learning and AI algorithms. Generative AI Examples: Simulating writing styles, generating images from text prompts, creating music compositions Predictive AI Examples: Financial forecasting, fraud detection, healthcare predictions, marketing strategies.

5 Generative AI Concerns

Hallucinations: AI can predict incorrectly due to biased/incomplete data or flawed models. Data Security: Risks when sharing proprietary data during model fine-tuning or processing sensitive requests. Plagiarism: AI models might replicate styles learned from public data, potentially leading to copyright issues. User Spoofing: AI can create realistic online profiles, challenging the identification of bot networks. Sustainability: High energy consumption during training increases the carbon footprint, though processing requests post-training is energy-efficient.

Inclusive Design

Inclusive Design is designing with diverse ways for people to participate, emphasizing the expertise of those who've faced exclusion. Universal Design focuses on a single design solution for everyone. Mismatches are instances when an individual's needs aren't met by the physical or digital world around them, leading to feelings of exclusion. These mismatches often arise from the way products or experiences are designed. Three Inclusive Design Principles: Recognize exclusion. Learn from diversity. Solve for one, extend to many.

Machine Learning

Machine learning gives computers and machines access to data (information), so they can then learn for themselves without a human having to program, type in or speak a command.

Natural Language Understanding/Natural Language Generation

Natural Language Understanding (NLU): Converts unstructured (human language) data to structured (organized) data. Interprets language to understand meaning & context. Natural Language Generation (NLG): Converts structured data to unstructured, enabling computers to generate human-like language. Can generate code (e.g., Python functions).

Natural Language Parsing

Natural language encompasses elements like vocabulary and grammar, and through parsing and NLP techniques, we can segment, identify root words, classify sentiments, and understand the intention and context behind words.

Einstein Discovery

Salesforce Einstein Discovery augments your business intelligence with statistical modeling and supervised machine learning in a no-code-required, rapid-iteration environment. Models are based on a comprehensive, statistical understanding of past outcomes that are used to predict future outcomes and suggest improvements. Einstein Discovery enables you to: -Identify, surface, and visualize insights into your business data. -Predict future outcomes and suggest ways to improve predicted outcomes in your workflows. Users can simulate scenarios to predict outcomes and suggest improvements. Salesforce offers notifications for data issues and bias detection to ensure ethical modeling.

Difference between structured and unstructured data

Structured data is organized with labels (e.g., spreadsheet) Unstructured data lacks a clear format (e.g., news article).

Supervised vs. unsupervised learning

Supervised learning uses labeled data where input has a known output Unsupervised learning finds patterns in unlabeled data. AI tries to find connections in the data without really knowing what it's looking for.

Values-Driven Design

Values-driven design ensures that organizational values guide product and service creation. Diversity, Equity, and Inclusion (DE&I) should be integrated into the design process, improving market potential. A "design with" approach involves pairing design experts with subject-matter experts to create inclusive products. Diverse perspectives are crucial internally for brand, policy, product, and service development.

Salesforce Generative AI Guidelines (Ethical Considerations)

(A.S.H.E.S.) Accuracy: Ensure reliable results; highlight uncertainties; maintain human review for critical tasks. Safety: Mitigate biases and harmful outputs; protect privacy and maintain data integrity. Honesty: Respect data sources; be transparent about AI-generated content. Empowerment: Balance between AI automation and human intervention; enhance human capabilities. Sustainability: Focus on right-sized models to optimize performance and reduce environmental impact.

5 Trusted AI Principles (Ethical Considerations)

(R.A.T.E.I.) Responsible: Safeguard human rights, protect data, enforce policies against abuse. Accountable: Hold selves accountable to customers, partners, and society; seek feedback for improvement. Transparent: make it clear how and why AI makes decisions. Data is controlled by customers. Empowering: Augment human decision-making; offer AI education and tools for responsible use. AI benefits should be accessible to everyone. Inclusive: Ensure AI represents diverse values; build with diverse data sets and teams. AI should improve the human condition and represent the values of all those impacted, not just the creators.

Cross Channel Behavioral Messaging

- Behavioral messaging is guided by consumer-expressed interests and triggers when a consumer takes specific actions, like clicking on a product. -Ethical considerations include user consent, transparency, and value addition. Overstepping boundaries risks eroding consumer confidence and brand trust. -Best practices include respecting preferences, targeting based on interests rather than demographics, and frequency capping to avoid overwhelming the consumer.

5 Stages AI Ethics Maturity Model

1. Ad hoc: Early stage, individuals recognize issues and advocate for ethical considerations. Informal discussions, internal social media, and tech talks spread awareness. Advocates can transition to full-time roles to establish an ethical AI practice. 2. Organized and Repeatable :Executive support secured, ethical principles established. Ethical AI team formed, prioritizing diversity in experience and background. Introduction of formal ethical reviews during product development. Concept of "ethical debt" (similar to "technical debt") is recognized. 3. Managed and Sustainable: Ethics integrated into product life cycle, formal processes like consequence scanning and model cards are introduced. Tools for fairness, accountability, transparency, and explainability (FATE) are utilized. Emphasis on real-world monitoring, considering global and cultural contexts, and expanding team globally. 4. Optimized and Innovative: Holistic, company-wide ethical approach, potentially transitioning to hybrid models for ethical oversight. Ethical debt is prioritized, with established metrics guiding product launches. Set minimum thresholds for launch in order to block the launch of any new product or feature that does not meet that threshold.

What are the 5 Einstein Products?

1. Einstein Bots: Utilize NLP to offer instant help, answering common questions or directing customers to the appropriate agent. 2. Einstein Prediction Builder: Point-click wizard for custom predictions that enables business predictions based on objects and fields. Ex: Will this customer (object) attrit (field)? 3. Einstein Next Best Action (NBA): Provides intelligent, context-aware recommendations using rules-based and predictive models. 4. Einstein Discovery: Predicts outcomes and uncovers insights on data patterns to understand customer behaviors. 5. Einstein with Generative AI: Leverages generative AI to produce hyper-relevant content, enhancing personalization and customer experiences.

Neural Networks

A neural network is a method in artificial intelligence that teaches computers to process data in a way that is inspired by the human brain. Neural Networks use weights and biases to enhance AI model predictions. They comprise of input nodes connected through various nodes (scenarios) to produce an output. Neural networks improve over simplified models by considering multiple scenarios and adjusting for complexities in data.

Bias in AI

AI systems rely heavily on data, which can introduce various biases, including measurement, association, and societal biases. While there's a difference between ethical and legal standards, both are crucial in AI development. The introduction of bias can come from assumptions, training data, model design, or human intervention. Importantly, unchecked biases in AI can be amplified in outcomes. Types of Bias: Measurement Bias: Incorrect data labeling/categorizing. Type 1 vs. Type 2 Error: False positives vs. false negatives. Association Bias: Stereotype-based labeling. Confirmation Bias: Reinforcing pre-existing ideas. Automation Bias: Imposing system's values on others (e.g. beauty contest judged by AI). Societal Bias: Reflects past prejudice towards marginalized groups (e.g. redlining). Survivorship Bias: Focuses on those who "survived" a process. Interaction Bias: Biased results due to human interaction. How Bias Enters System: Assumptions: Creator's preconceptions about system design. Training Data: Biased data influences AI learning. Model: Inherent biases in chosen factors for training. Human Intervention: Editing or neglecting training data.

Data & Algorithm Bias Removal

Addressing bias in AI is both a technical and social challenge. Use proactive strategies, like premortems, to anticipate issues and continually evaluate training data to ensure unbiased results. Consider the societal influences on data, and prioritize ethical practices, valuing human impact and inclusivity, in AI development and application. Premortem: A proactive approach to identify possible issues before they occur. Encourages open discussions about potential concerns. Dataset Biases: Ensure representation, avoid generalizations. Address "unknown unknowns" (highly confident but wrong predictions). Regularly Check Training Data: Address quality issues early. Monitor released models for evolving biases and cultural shifts. Engage Communities: Involve impacted communities in review processes. Allow users to correct/opt out of data. Ethical AI: It's a combination of technical & social efforts. Diverse teams reduce biases. Prioritize human values and responsible design/use.

Data Managements Best Practices

Build a Data Management Strategy: A clear data strategy helps keep your organization on track. Improve Data Quality: Tracking, reporting, and the effectiveness of your Salesforce deployment depend on getting clean data Import Data: Bring your existing data into Salesforce so you can include past records in your tracking and reporting. Maintain and Clean Up Data: keep all of your data clean in the long term

California Privacy Rights Act (CPRA)

CPRA expanded the privacy and data protection obligations under the CCPA, making it both more comprehensive and aligned to global standards. Any Business with an annual revenue of $25 million dollars, handles the data of 100,000 California consumers or devices, Derives 50% or more of its revenue from Selling or Sharing Personal Information must comply. Introduces new terms such as "Contractor" and "Sensitive Personal Information", adjusts the thresholds for compliance, and presents additional obligations akin to the EU's GDPR. The CPRA also enhances individual privacy rights and establishes a new enforcement body, the California Privacy Protection Agency (CPPA).

Chatbot

Chatbots are applications that mimic human conversation, allowing customers to engage with a computer rather than a human representative. They can enhance customer relations and digital engagement. While not all chatbots use AI, many are powered by natural language processing (NLP) or natural language understanding (NLU) technologies like Salesforce Einstein, which train bots to give human-like, automated responses. Benefits include quick answers, reduced wait times, and efficient redirects for customer inquiries. To implement an effective chatbot, careful planning is essential. Note: Service Cloud License from Salesforce is needed to set up a chatbot

Relationship Design Mindset

Compassion: Lead by connecting and valuing humanity; practice by rewarding curiosity and acknowledging others' experiences. Courage: Push beyond comfort zones, advocate for values; inspired by Salesforce CEO's ethos about improving the world. Intention: Focus on trust over transactions; design with relationship goals and be aware of consequences. Reciprocity: Emphasize two-way relationships and continuous feedback; co-create and gather insights. Outcomes: Enhanced value exchange, stronger business ties, positive societal and environmental impact. RD for Business Roles: Marketers: Value engagement over impressions. Salespeople: Discover needs via conversation, not just pushing products. Designers/Engineers: Focus on relationships rather than just features.

Ethics by Design

Create tech with positive intent and minimize negative impact. It must be designed, developed, and used in an ethical and humane way. Salesforce emphasizes the ethical and humane use of technology, driven by its core values of Trust, Customer Success, Innovation, Equality, and Sustainability. They've established the Office of Ethical and Humane Use to oversee product impact, embed ethical principles in product design, and collaborate with organizations for ethical advancement. Their guiding principles focus on human rights, privacy, safety, honesty, and inclusion. SF Ethical Differentiators: -AI Tools like the Einstein platform prioritize fairness, transparency, and inclusivity. -Language inclusivity is emphasized, with non-inclusive terms replaced. -Ethical guidelines have been developed specifically for products in response to challenges like the COVID-19 pandemic.

Cycle of Exlusion

Cycle of Exclusion: A framework to understand and counteract exclusion in design. Consists of five elements: Why we make (Motivation) Who makes it (Designer) How we make (Methods & Resources) Who uses it (User Assumptions) What we make (Product/Solution) Shut-In-Shut-Out Model: A traditional way of thinking about inclusion. It visualizes inclusion as "shutting in" a certain group and excluding others. This model promotes fixed thinking about who's "in" and "out". Exclusion Habits: Practices that lead to or perpetuate exclusion. By challenging these habits, we can identify and resolve mismatches in design. Risk of a Hero Complex: Avoid making assumptions about user needs to "save the day". Base design on real input rather than stereotypes. Key Takeaway: Inclusion in design is crucial. It's better to incorporate early on than to retrofit. Avoid exclusion by questioning our design habits and promoting inclusive practices.

Individual Rights under GDPR

Data Access: Individuals can confirm if their data is being processed and request specifics about such processing. Right to Object: Individuals can object to their data being processed, especially for direct marketing. Data Rectification: Individuals can ask for correction or completion of inaccurate or incomplete data. Restriction of Processing: Individuals can request limited access to and modification of their data. Data Portability: Individuals can request their data in a common format to transfer to another company. Right to Erasure: Individuals can request deletion of their data under specific conditions, also known as "the right to be forgotten".

How to Improve Data Quality with SF

Data Management Plan Components: Naming: Standardize naming conventions. Formatting: Standardize date and money representation. Workflow: Process for record lifecycle. Quality: Standards for data quality. Roles and Ownership: Accountability for data changes. Security and Permissions: Ensure data privacy. Monitoring: Quality control and dashboard metrics. Salesforce Implementation: Required Fields: Designate necessary fields. Validation Rules: Ensure correct data format. Workflow Rules: Automate internal procedures. Page Layouts: Custom layouts for relevance. Dashboards: Visual representation for business objectives. Data Enrichment Tools: Update data regularly. Duplicate Management: Prevent duplicate records. Custom Field Types: Standardize data entries.

Why is Data Quality Important?

Data quality is crucial for effective business operations and decision-making. Poor data leads to lost revenue, inefficiencies, and compliance challenges, while good data enables precise targeting, improved efficiency, and trust-building. Additionally, the integrity of AI predictions hinges on high-quality data. Good Data Benefits: Prospecting, increased efficiency, customer trust, and improved decision-making. Tools, such as the Data Quality Analysis Dashboards App from Salesforce AppExchange, can be used to review and enhance data quality attributes including age, completeness, accuracy, and consistency. The goal is to identify data gaps and implement improvements.

Deep Learning

Deep Learning involves adding extra layers to neural networks, enabling them to discern hidden patterns in data.

Inclusive Product Design

Design Context & Inclusion Designs are influenced by historical decisions. To combat exclusion, don't design based on assumptions. Co-design with experts from the targeted group. Misconception of Average "Average human" is a myth based Designing for the "average" can lead to exclusion; e.g., US Air Force jet cockpit design for "average" and fit no one. Data Interpretation Data, including big data, can be misleading. Thick data gives context, revealing the 'why' behind behaviors, complementing big data. Persona Spectrum & Inclusive Design Universal design = one-size-fits-all. Inclusive = one-size-fits-one. Persona spectrum identifies a range of user types, focusing on mismatches in interaction. E.g., Closed captioning was for the deaf but benefits many situations. Inclusive Motivation Aim isn't to fix a disability, but to enable fuller participation. Understand human motivations for inclusive design, not just functional needs.

Ecosystem Mapping

Ecosystem Mapping is a tool to identify and understand all key influencers of a product or service, including internal teams and leadership. Through stakeholder mapping, designers categorize individuals by their support levels using colors (red, yellow, green). The goal is to address challenges, align stakeholders, and recognize champions and detractors to successfully bring an idea to market.

Einstein Next Best Action

Einstein Next Best Action helps users make informed decisions by offering smart recommendations. It integrates various sources of business insights, allowing for unified data views, simplifying predictive models, and automating actionable recommendations. Provides a centralized hub for insights across Sales, Service, Marketing, and more. It allows for quick, context-sensitive decisions, improving efficiency and customer satisfaction. Independent software vendors can also customize it for specific business needs.

Ethical Use Principles and Best Practices in Personalization

Ethical Use of Technology at Salesforce: Salesforce emphasizes the ethical use of technology through its Responsible Marketing Principles. - Be transparent about security measures and data usage to build trust. - Obtain explicit consumer consent before data collection. - Personalize experiences based on customer intent, not just demographics, to avoid bias. - Offer clear benefits in exchange for data to ensure customer loyalty. Prioritize Behavior-Based Intent Over Demographic Attributes: - Relying solely on demographics can introduce bias and limit marketing reach. - Interest- and intent-based targeting is more effective and avoids reinforcing stereotypes. - Be cautious of indirect bias, such as correlations between postal codes and racial populations. - Focus on customers' behaviors and intentions rather than static attributes. - Using tools like Marketing Cloud Personalization Recipes can help target based on behavior instead of just attributes. Avoid binary identifiers, like gender, which can limit audience and alienate customers.

GDPR Key Principles

Fairness and Transparency: Organizations should process personal data lawfully and transparently, informing individuals about data collection and usage. Purpose Limitation: Personal data should only be collected for clear, explicit purposes and not be used beyond those intentions. Data Minimization: Only necessary and relevant personal data should be collected. Accuracy: Personal data should be up-to-date and accurate. Data Deletion: Data should only be retained as long as needed for its intended purpose. Security: Adequate technical and organizational measures, including encryption, pseudonymization, and anonymization, should be used to protect personal data. Accountability: Data controllers are responsible for GDPR compliance, incorporating principles like "privacy by design" and "privacy by default". Individual Rights: GDPR grants individuals rights like data access, the right to object, data rectification, restriction of processing, data portability, and the right to erasure. Organizations must operationalize these principles and be ready to address data subject requests for compliance.

Consequence Scanning Workshop

Framework adopted by Salesforce to ensure the ethical and humane use of technology. The process identifies six major unintended consequences, such as imbalances in technological benefits and environmental impact. Introduces "productive friction" into product development. Teams are prompted to reflect on the potential impacts of their creations. This method often takes the form of a workshop, aiming to build a "moral muscle memory" in product development. Consequence Scanning Workshops help teams assess the impacts of their designs on society by analyzing intended and unintended product consequences. The workshop consists of an Ideation Phase for deep thinking about consequences and an Action Phase for categorizing and addressing them. Key practices include focusing on specific features, involving diverse perspectives, and assigning preparatory task

EU Privacy Law (GDPR)

General Data Protection Regulation (GDPR) establishes rules for how companies, governments, and other entities can process the personal data of data subjects who are in the EU. Privacy is a Fundamental Right. The GDPR introduces stringent data processing guidelines, requiring clear consent, defining roles and responsibilities for data controllers and processors, and mandating breach notifications within 72 hours. It also enforces stricter fines, regulates data profiling, and expands individual data rights, with centralized enforcement for organizations operating across the EU.

Generative AI

Generative AI creates new content (text, images etc) using existing content it was trained with. Examples include text summarization, translation, error correction, question answering, guided image generation, and text-to-speech. Large Language Models (LLMs) like ChatGPT can have human-like interactions due to extensive training on vast data.

Human Centered Design (HCD)

HCD is a creative approach to problem-solving focused on understanding and addressing people's needs Core Components: Desirable: People need it. Feasible: It can be built. Viable: Fits the business model. Applications: User Experience (UX) Design: Focus on a person's experience with a product, aiming for ease of use, efficiency, and enjoyment. Example: airline emails you when it's time to check in for your flight. You click the link attached to the email, which launches the check-in webpage for your flight. It shows a large button that says, "Check In." Service Design: Considers all interactions a person has with a company, both digital and physical, aiming for ease, consistency, and quality of overall experience. Example: multiple touchpoints from online check-in, to airport kiosks, to in-flight services. Relationship Design: Builds on UX and service design, emphasizing connections between people, companies, and communities. Prioritizes engagement, connection, and social values. Also incorporates inclusivity, sustainability, and ethics. Examples: Booking for groups with auto-seat assignment, Non-binary gender options, customer community creation. Example: Airline travel can be enhanced by considering the needs and experiences of individuals, fostering connections, and ensuring sustainable and ethical practices.

Generative AI Development & Ecosystem

Key Drivers of Generative AI: -Huge amounts of training data, e.g., billions of web pages. -Advancements in neural network architecture, especially the transformer model. -Enhanced computational power, benefiting from parallel computing. AI Tech Stack Layers: -Compute hardware providers (e.g., Nvidia). -Cloud platforms (e.g., Google, Amazon, Microsoft). -AI models like LLMs (e.g., GPT4, Claude). Infrastructure optimization: curated datasets, analytics, and fine-tuning tools. Developers can access existing LLMs via API. -Applications: LLM-powered standalone tools & plugins.

AI Classification

Machine learning that involves assigning labels to data points. Labels can be discrete or continuous. Goal is to create a model that can accurately predict the label for new data points. Examples include Credit Scoring: Predicting whether a person will default on a loan (binary classification: "will default" vs. "won't default") or Email Filtering: Classifying emails as "spam" or "not spam"

3 Business Areas that Benefit from AI

Marketing: -Personalizes customer outreach through predictive behavior. Optimizes message sending time and frequency. -Uncovers deep market insights for improved communications. Sales Productivity: -Prioritizes sales leads based on conversion likelihood. -Auto-updates CRM data and recommends products for up-sell. -Assists sales managers with predictive revenue forecasting. Customer Service: -Automates email classification and routing to appropriate teams. -Provides real-time solutions for agents during live calls. -Employs chatbots for self-service support.

Rights under the California Consumer Privacy Act (CCPA)

Right to Notice: Clear information about what Personal Information is collected and why. Right to Access: Consumer ability to see categories, sources, and specific pieces of their personal data. Right to Opt Out (or In): Ability to refuse sale of personal data; special rules for children exist. Right to Request Deletion: Can ask businesses to delete personal data, with exceptions. Right to Equal Services & Pricing: No discrimination if CCPA rights exercised. A Business cannot reduce the quality of services or charge a higher price to a Consumer exercising their rights under the CCPA. Key Compliance Requirements for Businesses: Honor Consumer Rights: Respond to various consumer requests, ensuring identity verification and analyzing applicability. Provide Multiple Contact Methods: At least a toll-free number; can also provide email, postal address, website link. Transparency and Disclosure: Detailed public declarations about data practices; annual updates. Consent: "Do Not Sell My Personal Information" link on websites; opt-in for under 16s. Train Employees: Ensure staff is informed about CCPA and can assist consumers. Vendor Contracts: Ensure Service Providers have strict contracts aligning with CCPA.

Sales Cloud Einstein

Sales Cloud Einstein, an affordable Salesforce solution that democratizes AI for sales teams. It promises to offer all the benefits of a high-cost AI system without the need for additional staff or complex setups. Examples: Einstein Activity Capture: Automatically logs sales activities, reducing manual data entry. Connects with Gmail, Microsoft Office 365, etc. and shows activities directly in Salesforce. Addressed concerns:Security: Controls over which activities are added and visibility settings.Reporting: Advanced analytics dashboards and reports available. Einstein Automated Contacts: Finds and suggests or automatically adds new contacts based on email and event activity. Allows reps to stay focused on selling. Einstein Lead Scoring: Uses AI to analyze the history of lead conversions, revealing hidden patterns. Scores leads based on how well they match specific lead conversion patterns: high scores indicate a high potential lead. Provides clear reasons for each lead score, allowing reps to understand and trust the AI's recommendations. Adjusts its analysis continuously as new lead conversion data becomes available, ensuring scores remain accurate. Offers operational and analytics dashboards to monitor the effectiveness of the lead scoring, correlating lead scores to conversion rates and business impact. Einstein Opportunity Scoring: Works similarly to Lead Scoring by assessing past opportunities to predict future ones. Each opportunity is given a score backed with contributing factors. Einstein Opportunity Insights: Uses machine learning and sentiment analysis to aid sales reps. Offers smart predictions, reminders, and actionable insights. Insights are specific and tailored to an organization's patterns and data. Einstein Forecasting Features: utilizes artificial intelligence to provide sales managers with predictions about their team's future performance. The tool displays a prediction graph showing past opportunities and future forecasts. It offers key performance indicators such as the Einstein Prediction, Einstein Prediction to Quota Gap, and Closed to Quota Gap. The Forecasts Page provides detailed insights for each sales rep and team, and the Einstein Prediction Details Panel delves deeper int

Jobs to Be Done Framework (JTBD)

The Jobs to Be Done (JTBD) framework guides organizations in aligning product/service design with customer goals. It emphasizes understanding both functional and emotional needs. The four core principles are: Customer-centric: Prioritize customer conversations to identify jobs. Solution Agnostic: Focus on jobs, not just existing solutions. Stable Over Time: Address long-term, consistent customer needs. Measurable Outcomes: Ensure job effectiveness is quantifiable.

CCPA Best Practices

Update Privacy Notices & Disclosures: Align with GDPR changes from 2018.Incorporate "Do Not Sell My Personal Information" links for businesses selling data. Regularly update policies due to evolving interpretations. Handle Consumer Requests: Provide an easy system for data access requests. Verify consumer identity before sharing data. Respond within 45 days of request. Manage Data Sale Opt-Outs: Place a conspicuous opt-out link on websites. Carefully manage data to prevent unwanted sales after opt-out. Monitor definitions and interpretations of "Sale" under CCPA. Prevent Data Breaches: Apply reasonable security based on data sensitivity. Adopt a risk-based approach to identify and mitigate security vulnerabilities. Salesforce's Role: Salesforce functions as a Service Provider under CCPA. Provides assistance to customers for CCPA compliance. Existing Data Processing Agreement (DPA) aligns with CCPA, with a new DPA available for those who want a specific CCPA focus.

5 W's of Data Quality

Where: Understand data sources, filters, and categorization. When: Ensure data freshness. Outdated data can mislead predictions. Who: Data should represent all users. Check for skewed distributions. What: Be wary of biased variables (e.g., race, gender) and their proxies. Special Note: Special Data Categories: Avoid biased decision-making by being cautious about using sensitive data like race, gender, etc. Proxy Variables: Variables indirectly representing a biased category (e.g., zip code + income = race). Data Leakage: Using data known after the event of interest (e.g., using umbrella usage to predict rain).