Accelerate Your Marketing Innovation with Einstein
- Effortless Einstein Generative AI Activation: Instantly enable Einstein generative AI to access advanced features like automated segment creation and AI-powered content generation.
- Agentforce for Creative Campaigns: Leverage Agentforce to draft engaging marketing content, streamline campaign creation, and boost creativity. Build custom agents for reporting, content generation, journey creation, and more.
- Einstein Segment Creation: Rapidly build audience segments using intuitive natural language prompts, making segmentation faster and more accessible for marketers.
Key Advantages
- Enhanced Personalization: AI-driven segmentation and content recommendations.
- Advanced Reporting & Optimization: Real-time analytics for campaign performance.
- Einstein Integration: Automated insights and generative AI for smarter decision-making.
- Streamlined Setup: Faster onboarding and configuration, reducing reliance on technical teams.
Bottom Line: Marketing Cloud Next empowers marketers to deliver tailored experiences at scale, backed by robust analytics and AI.
Preparing for Einstein Generative AI
Before enabling Einstein Generative AI, it's critical to understand the implications for data privacy and security:
- Data Handling: Einstein generative AI is designed with trust as its top priority. Salesforce’s architecture ensures your data is secure and not used to train public models.
- Data Masking: Sensitive information must be protected from exposure to large language models (LLMs). The Einstein Trust Layer provides customizable data masking policies.
- Compliance: Ensure your setup aligns with regulations like GDPR, HIPAA, and CCPA.
Tip: Review your organization’s privacy and compliance requirements before activating Einstein features.
Einstein Trust Layer: Data Masking Explained
The Einstein Trust Layer uses data masking to safeguard sensitive information before it reaches the LLM. There are two main approaches:
-
Pattern-Based Masking
- Uses machine learning, regex, and context words to automatically identify and mask sensitive data in prompts.
- Supports data types like names, emails, credit cards, phone numbers, company names, passport numbers, SSNs, and more.
- Languages supported: English, French, German, Italian, Japanese, and Spanish.
- Limitations: Less predictable; may misclassify data (e.g., sales order numbers vs. phone numbers).
-
Field-Based Masking
- Relies on Salesforce metadata to manually tag and mask sensitive fields.
- More predictable; applies to any Salesforce field.
- Limitations: Limited to merge fields in prompts referencing record fields and related lists.
Best Practice: Use field-based masking for predictable protection, and pattern-based masking for broader coverage.
Safety and Security Features
-
Harmful Content Detection
- Toxicity Detection: Prompts and responses are evaluated for abusive, violent, or harmful content. Responses are scored and logged in the audit trail.
- Beta Features: Some detection capabilities (e.g., prompt toxicity, prompt injection) are in pilot/beta and subject to Salesforce’s Beta Services Terms.
-
Security Policies
- Prompt Injection Detection: Identifies attempts to manipulate model behavior via crafted prompts.
- Audit Trails: All detection events are logged for compliance and review.
-
Data Collection and Storage
- Data Cloud Integration: Einstein generative AI data can be collected and stored in Data Cloud for analytics, prompt optimization, and LLM training.
- Privacy Controls: Ensure only necessary data is collected, and that it is masked/encrypted as required.
Note: Always review beta features’ terms and use them at your organization’s discretion.
Steps to Enable Einstein Trust Layer
- Review privacy and compliance requirements.
- Classify and tag sensitive Salesforce fields.
- Configure field-based and pattern-based masking policies.
- Enable safety and security features (toxicity, prompt injection detection).
- Integrate with Data Cloud for analytics, ensuring privacy controls are in place.
- Monitor audit trails and update policies as needed.
Final Thoughts
Enabling Einstein Generative AI in Marketing Cloud Next unlocks powerful personalization and optimization capabilities. However, it’s essential to configure the Einstein Trust Layer thoughtfully to protect sensitive data, comply with regulations, and maintain customer trust. By following the steps above, you can confidently harness AI while keeping your organization’s data safe.