Duplicate Contact Reconciliation Agent

Tarabiscodé

See how the Duplicate Contact Detector uses Superhuman Go and Superhuman Docs to identify, review, and safely reconcile duplicate contacts.

The agent runs every Monday, compares contact information, calculates similarity scores, and generates clear reconciliation recommendations. A person reviews each match and approves or rejects it before any data is modified.

This demo covers the complete workflow: detect, review, decide, and reconcile. The template can also be adapted to other duplicate-record use cases.

Superhuman Docs template:

https://docs.superhuman.com/d/_dfYBgk7YE8u/_subvU2e-

Duplicate Contact Detector agent:

https://go.superhuman.com/builder/a183d7f0-cb37-45f6-bbef-27d26b202cb0?agent=75085

Transcript

00:00
Hi, I'm Elise, and this is my duplicate contact reconciliation agent.
00:05
This superhuman agent finds potential duplicate contacts, prepares reconciliation recommendations, and keeps every final decision under human control.
00:15
So the workflow has 4 steps: detect, review, decide, and reconcile.
00:21
Here, the Sprint Docs table contains all the contacts that the agent analyzes.
00:27
So every Monday at 9 AM, the agent compares names, email addresses, phone numbers, companies.
00:33
It calculates a similarity score and identifies pairs that are likely to represent the same person.
00:43
Here you have the agent, and I'm gonna ask him to run.
00:50
The agent reads the contacts table.
00:54
It detects new potential duplicates every Monday 9 AM.
00:58
As I said, and it writes its findings back to the document.
01:05
So we can see that the agent detected 8 duplicate pairs.
01:12
So we can go back to the document and just review these duplicates.
01:21
So this view enable us to compare both contacts, review the similarity score.
01:31
For instance, here it's 95%, and we can just read the recommendation.
01:38
For instance, in this example, we can see that contact A would be retained with the first name and the last name, but the email would be— the goal is to keep the email of the second contact, the phone number of the first one, the company of the second, and the status from the first one, you see.
02:00
So you can either approve merge or reject.
02:05
So here I approve, and I can just go to the second duplicate, and I can again approve.
02:18
And then approve here all of these.
02:22
Approve.
02:24
And I'm done for that.
02:30
here I've asked the agent to reconcile the records that are marked as ready to merge.
02:37
So we can see that he has reconciled all the 8 records.
02:43
So he has done the modifications, the updates, and then archive all the 8 duplicates.
02:53
So today the process is kind of manual, this step is kind of manual, but we could create another agent that just detects and does the real-time reconciliation when you update, you know, the status.
03:13
So that's all, thank you very much and see you next time!
Duplicate Contact Reconciliation Agent — Tella