🤖
Module 4: Einstein Ecosystem (Tools)
Trust Layer, Prompt Builder and CRM Tools
⏱️ Estimated reading time: 20 minutes
Chapter 9: Einstein Trust Layer
It is the security architecture that allows using external LLMs (like OpenAI) without exposing company data.
Key Components:
1. Secure Gateway: Secure entry point.
2. Data Masking: Replaces sensitive data (emails, names) with generic text before sending to the LLM.
3. Zero Data Retention: Legal and technical agreement whereby the LLM provider DOES NOT save your data or use it to train their model.
4. Toxicity Detection: Scans the LLM response for harmful content before showing it to the user.
Key Components:
1. Secure Gateway: Secure entry point.
2. Data Masking: Replaces sensitive data (emails, names) with generic text before sending to the LLM.
3. Zero Data Retention: Legal and technical agreement whereby the LLM provider DOES NOT save your data or use it to train their model.
4. Toxicity Detection: Scans the LLM response for harmful content before showing it to the user.
🎯 Key Points
- ✓ The Trust Layer makes Generative AI safe for business.
- ✓ Your data NEVER trains the public model (Zero Retention).
Chapter 10: Generative and Predictive Tools
Generative Tools (Create):
- Einstein Copilot: Conversational assistant (chatbot for employees) integrated into the sidebar. Can perform actions (search records, summarize calls).
- Prompt Builder: Tool to create, test, and manage reusable prompt templates. Allows including dynamic data (Grounding).
Predictive Tools (Calculate):
- Lead Scoring: Scores potential customers.
- Einstein Bots: Traditional menu-based chatbots (not generative).
- Opportunity Forecasting: Predicts how much you will sell this month.
- Einstein Copilot: Conversational assistant (chatbot for employees) integrated into the sidebar. Can perform actions (search records, summarize calls).
- Prompt Builder: Tool to create, test, and manage reusable prompt templates. Allows including dynamic data (Grounding).
Predictive Tools (Calculate):
- Lead Scoring: Scores potential customers.
- Einstein Bots: Traditional menu-based chatbots (not generative).
- Opportunity Forecasting: Predicts how much you will sell this month.
🎯 Key Points
- ✓ Copilot = Chat assistant.
- ✓ Prompt Builder = Create templates.
- ✓ Lead Scoring = Predictive (not generative).
Chapter 11: Prompt Engineering
It is the art of writing effective instructions to get the best response from the LLM.
Elements of a good Prompt:
1. Role: Who should the AI be? (e.g., 'Act as a sales expert').
2. Task: What do you want it to do? (e.g., 'Write an email').
3. Context: Necessary background (e.g., 'The customer is angry about a delay').
4. Constraints/Format: Length, tone, style (e.g., 'Less than 100 words, formal tone').
Iteration: It is normal not to get the perfect result the first time. You must refine the prompt repeatedly.
Elements of a good Prompt:
1. Role: Who should the AI be? (e.g., 'Act as a sales expert').
2. Task: What do you want it to do? (e.g., 'Write an email').
3. Context: Necessary background (e.g., 'The customer is angry about a delay').
4. Constraints/Format: Length, tone, style (e.g., 'Less than 100 words, formal tone').
Iteration: It is normal not to get the perfect result the first time. You must refine the prompt repeatedly.
🎯 Key Points
- ✓ More context = Better response.
- ✓ Always define Role and Task clearly.