📊
Module 3: Data (The Fuel)
Data Quality, Privacy, Data Cloud and Grounding
⏱️ Estimated reading time: 30 minutes
Chapter 6: Data Quality
AI is only as good as the data that feeds it ('Garbage in, Garbage out'). Key Dimensions:
- Accuracy: Does the data reflect reality?
- Completeness: Are all necessary fields filled?
- Consistency: Is the format the same across systems?
- Timeliness: Is the data up to date?
- Duplicates: Repeated records that confuse AI and skew predictions.
- Accuracy: Does the data reflect reality?
- Completeness: Are all necessary fields filled?
- Consistency: Is the format the same across systems?
- Timeliness: Is the data up to date?
- Duplicates: Repeated records that confuse AI and skew predictions.
🎯 Key Points
- ✓ Duplicate data = Confused AI.
- ✓ Incomplete data = AI with bias or errors.
- ✓ Cleaning data is step 1 before implementing AI.
Chapter 7: Privacy and Security
PII (Personally Identifiable Information):
Any data that identifies a person: Email, Phone, ID, IP, Address. These data must NEVER be used to train public models.
Regulations:
You must know GDPR (Europe) and CCPA (California). Compliance is not optional.
Shared Responsibility Model:
- Salesforce secures the platform (the cloud).
- The customer secures their data (what they put in the cloud).
Any data that identifies a person: Email, Phone, ID, IP, Address. These data must NEVER be used to train public models.
Regulations:
You must know GDPR (Europe) and CCPA (California). Compliance is not optional.
Shared Responsibility Model:
- Salesforce secures the platform (the cloud).
- The customer secures their data (what they put in the cloud).
🎯 Key Points
- ✓ Protecting PII is priority number 1.
- ✓ The customer owns their data, not Salesforce.
Chapter 8: Preparation and Grounding
Data Cloud (Harmonization):
Tool that unifies data from multiple sources (Salesforce, external webs, data lakes) into a single customer profile in real-time.
Grounding:
The process of giving context to AI using *your* trusted data. Instead of letting AI invent, you say: 'Use THESE data from my database to answer'. This drastically reduces hallucinations.
Tool that unifies data from multiple sources (Salesforce, external webs, data lakes) into a single customer profile in real-time.
Grounding:
The process of giving context to AI using *your* trusted data. Instead of letting AI invent, you say: 'Use THESE data from my database to answer'. This drastically reduces hallucinations.
🎯 Key Points
- ✓ Grounding connects the generic LLM with your real data.
- ✓ Data Cloud solves the 'information silos' problem.