Audit messy CRM fields and create a cleanup plan with AI

Start with 5-8 fields that influence segmentation, routing, scoring, attribution, lifecycle reporting, or sales handoffs. Good candidates include lifecycle stage, lead status, lead source, industry, persona, company size, country, product interest, and closed-lost reason. Do not audit every field at once. Pick the fields that break decisions.

Export a sample of records from Salesforce with the selected fields, record IDs, owner, created date, last modified date, source, lifecycle stage, and relevant outcome fields. In Google Sheets, add columns for field name, current value, record count, sample records, dependent reports, dependent automations, and cleanup recommendation.

Paste the field-value export into Claude in batches and ask it to cluster values that mean the same thing. For example, 'Manufacturing,' 'Mfg,' 'Industrial,' and 'Discrete Manufacturing' may be variants or may need separate definitions. Ask Claude to propose canonical values, merge candidates, risky merges, and values requiring business-owner review.

Audit these CRM field values and recommend cleanup clusters.

Field name:
{{field_name}}

Field purpose:
{{field_purpose}}

Values with counts and samples:
{{field_values_with_counts}}

Known reports or workflows using this field:
{{dependencies}}

Output:
1. Recommended canonical values
2. Values that can safely merge
3. Values that are ambiguous
4. Values that should be retired
5. Values requiring business-owner review
6. Risk notes
7. Suggested validation rules

Do not assume similar words always mean the same thing. Flag uncertainty.

Review records where field combinations do not make sense: MQLs with no source, customers marked as open leads, closed-lost opportunities with no loss reason, enterprise accounts routed to SMB owners, or partner-sourced leads missing partner attribution. Add examples to the sheet and categorize the break as reporting, routing, attribution, scoring, or ownership.

Find likely CRM logic breaks in this record sample.

Records:
{{crm_record_sample}}

Field definitions:
{{field_definitions}}

Lifecycle and routing rules:
{{lifecycle_and_routing_rules}}

Identify records where field combinations suggest a problem. For each issue, output:
1. Record identifier
2. Suspected issue
3. Why it matters
4. Affected workflow: reporting, routing, attribution, scoring, ownership, or lifecycle
5. Recommended fix
6. Whether this requires manual review

In Google Sheets, create a decision table with each field issue, proposed action, affected record count, risk level, approver, system dependency, and rollback notes. Actions should include merge values, retire values, add validation, create missing value, update records, change routing logic, or leave unchanged. This turns the audit from interesting analysis into an executable cleanup plan.

Review the decision table with RevOps, sales leadership, marketing ops, and any team that owns dependent reporting. Confirm canonical values, retired values, validation rules, and records that require manual review. Do not run updates until each high-risk action has an owner and approval status.

Apply approved changes to a small sample first. Check whether routing, reports, lifecycle views, dashboards, and segmentation still behave correctly. Compare before-and-after record counts and confirm no active automations fired unexpectedly. Only then expand to the full cleanup set.