Feeding Customer Data to AI Models Without a Consent Framework
“Just because you CAN send customer data to an AI model doesn't mean you SHOULD.”
What Happened
Client wanted to use Einstein GPT to generate personalized marketing emails. We connected Data Cloud, pulled in customer purchase history, browsing behavior, and support interactions. Fed it all to the AI for email personalization. Nobody checked their privacy policy. Nobody reviewed consent records. Nobody asked legal. A customer in the EU filed a GDPR complaint because their support case details appeared in a marketing email that clearly came from AI. The DPA investigation took six months.
The Wrong Way
// Data Cloud + Einstein GPT - no consent check // Pull everything, ask questions later Data Cloud Data Stream: All customer interactions → Purchase history → Support case transcripts ← contains sensitive info → Website browsing behavior → Email engagement history Prompt Template: "Generate a personalized email for this customer using all available data about their history with us." // AI output: "We noticed you contacted support about [sensitive issue] // and wanted to follow up with a special offer..." // Customer: *files GDPR complaint*
The Right Way
// Step 1: Define consent model BEFORE connecting data
Individual Object (or Consent__c custom object):
→ Marketing_Consent__c: Boolean
→ AI_Processing_Consent__c: Boolean
→ Data_Retention_Consent__c: Boolean
→ Consent_Date__c: DateTime
→ Consent_Source__c: Picklist (Web Form, Email, Phone)
// Step 2: Filter data in Data Cloud based on consent
Data Cloud Segment: "AI Marketing Eligible"
→ Marketing_Consent__c = TRUE
→ AI_Processing_Consent__c = TRUE
→ Country != EU OR GDPR_Consent__c = TRUE
// Step 3: Prompt Template with data restrictions
Prompt: "Generate a personalized email using ONLY the following data:
- First Name: {!Contact.FirstName}
- Product Category Preferences: {!Contact.Preferred_Categories__c}
- Last Purchase Category: {!LastOrder.Category}
Do NOT reference:
- Support cases or complaints
- Specific purchase amounts
- Browsing behavior
- Any health, financial, or sensitive information"
// Step 4: Human review before send
// Step 5: Audit log of all AI-generated contentThe Lesson
Build a consent framework before connecting customer data to AI. Filter data based on consent flags. Restrict what the AI can reference. Log everything. Your legal team will thank you.
Enjoyed this? Get more like it.
Glen's Musings — AI, investing, and building things. Occasional. Free.
More AI & Agents Mistakes
Turning On Einstein Prediction Builder with Dirty Data
AI trained on garbage data gives you garbage predictions with confidence scores.
Read moreCareer-EndingDeploying Einstein Copilot Without Testing Its Guardrails
An AI assistant with full org access and no guardrails is a data breach waiting to happen.
Read moreAnnoyingBuilding Prompt Templates Without Grounding in Record Data
An ungrounded prompt hallucinates. A grounded prompt uses your actual Salesforce data.
Read more