Preparing Your Salesforce Org for AI: A Data Readiness Checklist

The pre-flight inspection that determines whether your AI takes off or crashes

The pattern repeats across every failed AI implementation I've seen:

1. Organization gets excited about AI capabilities
2. Team enables the feature quickly to show progress
3. AI produces embarrassing, inaccurate, or unusable results
4. Users lose trust and stop using the feature
5. Investment is wasted

The cause is almost always the same: the org wasn't ready. Not technically, not from a data quality perspective, not organizationally.

This checklist captures everything I check before enabling any AI feature in Salesforce. Use it to identify gaps before they become failures.

Section 1: Data Quality Assessment

AI accuracy depends entirely on data quality. Garbage in, garbage out.

1.1 Completeness Audit

For every object the AI will access:

Contact Records:

• Email populated on 90%+ of records
• Title populated on 80%+ of records
• Account relationship exists for 95%+ of records
• Phone populated on 70%+ of records

Account Records:

• Industry populated on 85%+ of records
• Address fields populated on 80%+ of records
• Website populated on 60%+ of records
• Employee count or size indicator on 50%+ of records

Opportunity Records:

• Amount populated on 95%+ of records
• Close Date populated on 100% of records
• Stage reflects actual status
• Primary Contact identified on 90%+ of records

Custom Objects:

• Identified critical fields for AI use
• Measured completeness for critical fields
• Gaps documented with cleanup plan

1.2 Accuracy Audit

Sample 50-100 records per object:

• Contact job titles match LinkedIn/company websites
• Account industry classifications are correct
• Phone numbers are valid and current
• Email addresses are deliverable
• Address data matches actual locations

Accuracy Threshold: 85%+ accuracy on critical fields before enabling AI

1.3 Consistency Audit

Check for standardization:

• State fields use consistent format (all 2-letter abbreviations or all spelled out)
• Phone numbers use consistent format
• Industry values follow single taxonomy (no overlapping categories)
• Title formatting is standardized (VP vs Vice President)
• Date fields use actual date type, not text

Run these reports:

• Group by State, count variations
• Group by Industry, identify duplicative categories
• Group by Title, identify similar titles with different formatting

1.4 Timeliness Audit

• 80%+ of Contact records updated in past 12 months
• 90%+ of Account records updated in past 12 months
• Activity data exists for recent time periods
• Opportunity stages reflect current reality (no stale "In Progress" opps)

Run this report: Records where Last Modified Date < TODAY() - 365

If more than 20% of records are over a year old, implement a refresh process before AI.

1.5 Duplicate Assessment

• Duplicate rules are active for Contacts, Leads, Accounts
• Historical duplicates have been merged
• Duplicate Record Sets are reviewed weekly
• Duplicate rate is below 5%

Check: Setup → Duplicate Management → Duplicate Record Sets

If significant duplicates exist, clean before enabling AI.

Section 2: Security and Access Configuration

AI must respect your security model.

2.1 Sharing Model Verification

• Organization-Wide Defaults are correctly configured
• Role Hierarchy reflects actual organizational structure
• Sharing rules cover necessary scenarios
• AI features will only access appropriate data

Test: For each AI feature, test with users at different levels. Verify they only see appropriate data.

2.2 Field-Level Security

• Sensitive fields identified (SSN, salary, health info, etc.)
• Sensitive fields have appropriate FLS restrictions
• AI features cannot access restricted fields inappropriately
• Trust Layer masking configured for sensitive fields

2.3 Object Permissions

• AI actions respect user object permissions
• AI cannot create/update/delete records users shouldn't modify
• CRUD permissions are appropriate for each AI use case

2.4 Trust Layer Configuration

• Trust Layer settings reviewed
• Data masking configured for PII
• Guardrails defined (topics AI shouldn't discuss)
• Audit logging enabled

Section 3: Technical Readiness

3.1 License Requirements

• Appropriate Einstein/Agentforce licenses available
• Licenses assigned to users who will use AI features
• License count sufficient for planned rollout

3.2 Feature Enablement

• Required features enabled in Setup
• Einstein features activated where needed
• Agentforce components configured
• Data Cloud connected (if required)

3.3 Integration Considerations

• AI features compatible with existing integrations
• External data sources accessible to AI where needed
• Integration users have appropriate permissions
• API limits sufficient for AI workload

3.4 Testing Environment

• Sandbox available for AI testing
• Sandbox has representative data (not production, but realistic)
• Test users configured with various permission levels
• Testing plan documented

Section 4: Governance Framework

4.1 Ownership

• Data Owner assigned for each AI-accessed object
• AI Feature Owner assigned (who's responsible for the AI?)
• Escalation path defined for AI issues
• Executive sponsor identified

4.2 Policies

• AI use cases documented
• Data quality standards defined for AI-critical fields
• AI guardrails documented
• Exception handling process exists

4.3 Monitoring

• AI performance metrics defined
• Dashboard created for AI monitoring
• Alert thresholds configured
• Review cadence established (weekly for new AI, monthly ongoing)

4.4 Documentation

• AI features documented for users
• Training materials available
• Known limitations documented
• FAQ prepared for common questions

Section 5: Organizational Readiness

5.1 Stakeholder Alignment

• Business stakeholders understand AI capabilities
• Realistic expectations set (AI augments, doesn't replace)
• Success criteria defined
• Failure scenarios discussed

5.2 Change Management

• Communication plan for AI launch
• Training schedule established
• Super users identified for support
• Feedback mechanism in place

5.3 Risk Assessment

• Potential AI failure scenarios identified
• Impact of each scenario assessed
• Mitigation strategies defined
• Rollback procedure tested

5.4 Legal and Compliance

• Legal review completed for AI use cases
• Compliance requirements identified (GDPR, HIPAA, etc.)
• Data Processing Addendums reviewed
• Customer communication required? Prepared?

Section 6: Rollout Plan

6.1 Phased Approach

• Pilot group identified (small, lower-risk use case)
• Success criteria for pilot defined
• Expansion criteria established
• Timeline for each phase

6.2 Pilot Configuration

• AI features configured for pilot scope
• Pilot users have appropriate access
• Monitoring active for pilot
• Feedback collection mechanism ready

6.3 Go/No-Go Criteria

• Data quality thresholds met
• Security configuration verified
• Governance framework in place
• Stakeholder approval obtained

6.4 Launch Readiness

• Training completed for pilot users
• Support resources available
• Known issues documented
• Day-1 monitoring plan active

The 30-60-90 Day Plan

Days 1-30: Foundation

Week 1:

• Run data quality audit
• Document gaps and issues
• Prioritize cleanup work

Week 2-3:

• Execute high-priority data cleanup
• Implement validation rules for critical fields
• Configure duplicate management

Week 4:

• Verify security configuration
• Configure Trust Layer
• Prepare sandbox for testing

Days 31-60: Configuration and Testing

Week 5-6:

• Configure AI features in sandbox
• Test with users at different permission levels
• Document findings and adjust

Week 7-8:

• Run pilot in sandbox with real scenarios
• Collect feedback and iterate
• Finalize configuration

Days 61-90: Pilot and Expansion

Week 9-10:

• Launch pilot in production
• Monitor closely
• Support pilot users

Week 11-12:

• Evaluate pilot results
• Adjust based on feedback
• Plan expansion if pilot successful

Red Flags That Should Pause AI Enablement

Stop and address these before proceeding:

Data Quality:

• More than 30% of critical fields are blank
• Accuracy audit shows less than 70% correct data
• More than 15% of records are duplicates
• More than 40% of records haven't been updated in a year

Security:

• No one can explain the current sharing model
• Sensitive data is accessible to too many users
• Trust Layer hasn't been configured

Organizational:

• No clear owner for the AI initiative
• Stakeholders expect AI to "just work" without data prep
• No budget for data cleanup
• No plan for handling AI failures

If any of these are true, fix them before enabling AI. Otherwise, you're setting up for failure.

The Readiness Score

Score each section:

• 0: Not started
• 1: Partially complete
• 2: Mostly complete
• 3: Fully complete

Section Scores:

• Data Quality (1.1-1.5): ___ / 15
• Security (2.1-2.4): ___ / 12
• Technical (3.1-3.4): ___ / 12
• Governance (4.1-4.4): ___ / 12
• Organizational (5.1-5.4): ___ / 12
• Rollout Plan (6.1-6.4): ___ / 12

Total: ___ / 75

Interpretation:

• 60-75: Ready for AI pilot
• 45-59: Address gaps before proceeding
• Below 45: Significant work needed; don't enable AI yet

Next Steps

1. Print this checklist
2. Score your current state honestly
3. Identify the biggest gaps
4. Create a remediation plan with timeline
5. Don't enable AI until you've addressed critical gaps

If you're planning an Agentforce implementation and want help assessing your readiness, Clear Concise Consulting offers AI readiness assessments. We'll identify gaps, prioritize remediation, and help you build the foundation AI requires.