Use Claude or Codex to detect duplicate companies, stale contacts, broken associations, and Salesforce or HubSpot mismatches, then review safe fixes in Sanka.
Last updated: May 16, 2026
This guide shows how to run CRM data cleansing with Sanka. Start by asking Claude or Codex to find issues and prepare recommendations, then review and approve fixes before records are merged, normalized, reassigned, or synced.
Claude/Codex
Scan Sanka, HubSpot, and Salesforce CRM data for duplicate companies, stale contacts, missing domains, owner mismatches, and broken associations. Prepare recommendations only; do not merge or update records yet.
Preparing CRM cleansing queueI found duplicate candidates, missing fields, stale contacts, and ownership mismatches. Please review the recommended fixes before applying any changes.
Ask for another cleansing check...
Before you start
Check that you have the following ready.
Sanka has the companies, contacts, deals, and associations you want to review
HubSpot or Salesforce is connected if you want cross-system checks
Your team has source-of-truth rules for fields such as owner, lifecycle stage, domain, address, and industry
You know which fixes can be auto-applied and which require review
You have permission to update, merge, and sync CRM records
Ask AI to find data issues
Ask Claude or Codex to produce a review queue rather than making changes immediately.
Sample prompt
/sanka Find CRM data issues across companies, contacts, deals, and associations. Group them by duplicate candidate, missing required field, stale owner, broken association, and CRM mismatch. Show recommended fixes and risk level. Do not apply changes yet.
Common checks include:
Companies with the same domain, name, phone number, or billing address
Contacts without email, company, role, or consent status
Deals assigned to inactive owners or mismatched companies
HubSpot and Salesforce fields that disagree with Sanka
Associations that point to archived or duplicate records
Review and approve fixes
Review high-confidence fixes first. Exact matches can often be applied in bulk, while fuzzy matches should stay in the queue until a person confirms them.
Sample prompt
/sanka For these duplicate company candidates, explain why each pair is likely or unlikely to be the same company. Recommend which fields to keep from each source, and keep all changes pending for review.
Before applying a fix, check:
Which record will remain as the primary record
Which values are kept, replaced, or left unchanged
Which downstream records are affected
Whether HubSpot or Salesforce should receive the update
Whether the change can be reversed from audit history
Apply and sync approved changes
After approval, apply the changes in Sanka and sync back to connected systems when needed. Keep sensitive or ambiguous changes out of auto-apply rules.
Checkpoints
Use Sanka logs to confirm what changed, who approved it, and which connected systems were updated.
Search Sanka...
Review CRM cleansing activity
Logs
Search logsAll actionsAll dates
ID / ActionDateTarget / ItemChangeActor
4HubSpot sync prepared2026/05/12 16:20Acme Inc. / CompanyPrepared updated domain and owner for syncClaude / Codex
