What it does
On record creation, finds similar existing records in real time, computes an AI match-confidence score, auto-merges high-confidence pairs, flags medium-confidence pairs for review, and ignores low-confidence matches.
You get
duplicate prevention at the point of entry
not a cleanup project later
How it works
Check at the moment of creation
Create WorkflowRuns the instant an account or user is created, so duplicates are caught before they spread rather than cleaned up months later.
Look for a match
configure_find_records_stepSearches for records that look the same: the same domain for accounts, the same email for people, a close name match, or the same primary contact, returning candidates with a similarity score.
Stop early if it is new
configure_workflow_branch_stepNo candidates means the record is genuinely new and it goes on to normal onboarding and enrichment. Candidates mean it carries on to scoring.
Score each pair
configure_ai_research_stepEvery candidate is scored 0 to 100 against the new record on all its fields, recent activity and context such as arriving from the same campaign. It knows two addresses at one company are usually two people rather than one duplicate, and it explains every score.
Act on the confidence
configure_workflow_multi_split_stepAbove 90 merges automatically. Between 60 and 90 goes to the review queue and is never merged on its own. Below 60 carries on as a new record, tagged for a later look.
Merge into the older record
configure_update_attribute_stepThe existing record survives so its history is kept, any new fields are copied across, and the new record is marked as merged into it. References to the new one still work and point at the survivor.
Queue the uncertain pairs
configure_create_task_stepA RevOps task showing both records side by side with the reasoning for the suspicion and a merge or separate decision, to be cleared within five working days before the queue builds up.
Publish and audit the merges
publish_workflowValidated and published, watching auto merges per week, the false positive rate from a sample audit of them, and the depth of the review queue, which grows when too much lands in the middle band.
Ready to run this recipe?
Tell Blu what you need. It builds the segment, content, journey, and dashboard in minutes.
Sign up free
