The problem this solves
Every CRM accumulates rot at a rate roughly proportional to how fast the business once grew. Addresses bounce because people moved on, whole domains die because companies folded or rebranded, and role accounts like info@ and sales@ sit in the contact table pretending to be people. None of it announces itself. The rot surfaces as symptoms: campaign bounce rates creeping up, sender reputation sagging, and counts on dashboards that everyone quietly discounts because they know the base includes ghosts.
The reason the sweep never happens is that deleting CRM data feels dangerous and checking it feels endless. Nobody wants to be the person who purged a record that mattered, so the safe-feeling choice is to leave everything, and the database becomes an archaeology site. Meanwhile the cost compounds invisibly: every send to a dead address chips at deliverability, every ghost row distorts a metric, and every list pull starts with silent mental subtraction.
The resolution is quarantine rather than deletion, applied on evidence rather than age. Verify every address, check every domain, classify the role accounts, and move what fails into a quarantined state where it stops polluting sends and counts while remaining fully recoverable. Then put the before-and-after in writing, because a hygiene process nobody can inspect is a hygiene process nobody will trust.
How the mission runs
- Inventory the contact base. The agent pulls the contact table from Zoho CRM and establishes the baseline: total records, apparent duplicates, and the distribution of address types. This starting snapshot anchors the final report, so the sweep's effect is stated in numbers instead of adjectives.
- Verify every address. Email Verification runs across the base, classifying each address: valid, bounced, catch-all, unverifiable, or role-based. Domains are checked too, so contacts at dead or parked domains are caught even where the mailbox check alone is inconclusive. Each classification is stored with the evidence that produced it.
- Quarantine on evidence. Rows that fail - hard bounces, dead domains, role accounts per your policy - are moved to a quarantined state in Zoho CRM rather than deleted. Each quarantined record notes why it was flagged and on what evidence, and the operation is idempotent, so re-running the sweep never double-processes a record.
- Preserve the judgment calls. Catch-all and unverifiable addresses live in a gray zone where evidence is genuinely ambiguous, so they are reported as a separate category with a recommended treatment rather than quarantined automatically. You decide the policy once, state it in the prompt, and future runs apply it consistently.
- Email the before-and-after report. The mission closes by sending the quality report through Zoho Mail: base size before and after, counts per failure category, the quarantine list itself, and the gray-zone cases awaiting your ruling. It is the document that turns a background cleanup into an inspectable, repeatable process.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A Zoho CRM where the ghosts are contained: bounced addresses, dead domains, and role accounts moved to quarantine with a reason recorded on every affected record, valid contacts untouched, and nothing deleted. In your inbox, a before-and-after quality report with counts per failure category, the full quarantine list, and the ambiguous cases held for your ruling, giving you both a cleaner base and the paper trail that proves it.
Make it yours
- Run it quarterly and keep the reports; the failure-rate trend across runs tells you how fast your base decays and whether acquisition sources differ in quality.
- Extend the policy to quarantine contacts with no activity beyond a threshold you choose, folding engagement decay into the same sweep and the same report.
- Add a restore pass: before each new run, re-verify the existing quarantine and surface any addresses that have come back to life, since domains and mailboxes do return.
Frequently asked questions
Does anything get deleted?
No. The sweep quarantines rather than deletes: flagged records are moved to a state where they stop appearing in sends and active counts while remaining fully recoverable. Deletion, if you ever want it, is a separate deliberate step you take against the quarantine list after reviewing the report.
What happens with catch-all domains?
They are reported as their own category instead of being silently quarantined or silently kept, because the evidence for them is genuinely ambiguous. The report recommends a treatment, you set the policy once, and every subsequent run applies your ruling consistently and notes it on each affected record.
Can I see why a specific record was quarantined?
Yes, in two places. The record itself carries a note stating the failure category and the evidence, and the emailed report lists every quarantined contact with the same reasoning. Every classification traces to the verification call that produced it, so any single quarantine decision can be audited in seconds.