Automate a weekly CRM clean-up list for duplicates and gaps

Every form, import and manual entry can create a second copy of the same person. Over months, reports double count, two reps contact the same lead, and fields such as owner or lifecycle stage go empty. Some CRMs help: HubSpot, for example, compares records daily to surface potential duplicates and lets you merge or reject them individually or in bulk. A weekly review of what remains still catches what built-in rules miss.

You add the job and connect read access to your CRM through a remote MCP server (Beta) you set up. Each week a scheduled agent reads contacts and companies changed in the last seven days, and looks for matching emails, near-identical names at the same company, and records missing the fields you list. Because it reads the records that changed in the last seven days, duplicates where neither record changed this week are not caught, so run a full pass by hand now and then. For each likely duplicate it shows both records side by side with the reason they match. It reads and reports. It merges and edits nothing.

The COO agent, as department head, reviews the list before it goes out: do the matches share a real identifier, are any two people with the same name at a large company wrongly paired, are the missing-field counts right. Then you approve, edit or skip it, and nothing is changed in your CRM until you do. Merging cannot always be undone, so keep every merge as your own decision.

Tell the agent which fields matter. A contact without an owner is a different problem from one without a phone number. Start with exact email matches, which are safe, and add fuzzy name matches once you trust the list. Keep a note of pairs you rejected so they do not return each week.