Upsell proof · multi-source · not the first-project door

Lead Ops Gate

Nothing auto-sends.

Proof of a deeper, multi-source Lead Ops-style gate (Contra Step 3 / Notion + AI/API desk) when one inbox is not enough. Form JSON, CSV, or inbound rows → score → CRM + Tasks → first-reply draft. A human Approves before send. Not the $199 / first-project door.

Unsure? Start with the Feedback Ops audit/install — this desk is upsell proof only, for when one inbox is not enough. feedback-ops-copilot.vercel.app · verdelabs.cloud/work

Nothing auto-sendssent stays false — even after Approve

upsell proof only
not the $199 door

  1. 01

    Lead in

    Multi-source: form JSON, inbound email, or a CSV row lands on the desk. Fictional .example contacts only.

  2. 02

    Classify / score

    A rule engine stamps hot / warm / junk with reasons you can show a buyer. No paid model in this mock.

  3. 03

    Approve-before-send

    A first-reply draft waits on Approve / Reject. Junk is held. CRM + Tasks upsert either way.

  4. 04

    Human send

    A person copies the approved draft later. This mock never flips sent — even after Approve.

3-minute path

This is the deeper desk. If you only need one inbox, start with Feedback Ops.

  1. Scan the gallery — Maya (hot), Elena (warm), Aiden (junk). All sent: false.
  2. Compare spray-and-pray vs the HITL gate board.
  3. Open Maya Chen. Confirm CRM + Task upsert and the held draft.
  4. Approve, then Reject. The stamp stays sent: false.
  5. Ingest the sample CSV (empty row) or a bad payload. Check Retries.

Before / after

Spray-and-pray vs the HITL gate board

Same fictional contacts. Left is the unsupervised blast. Right is this deeper desk — scored, held, a human Approves, sent: false.

Before

Spray-and-pray outreach

One sequence. No score. Junk rides along. The bot already sent.

  • Maya Chen

    Same “just checking in” sequence as the list buy.

    Hot inbound treated like a cold blast.

    auto-sent

  • Sam Ortiz

    Auto-reply fired before anyone read “no auto-send.”

    Asked for coverage without unsupervised mail.

    auto-sent

  • Priya Nandakumar

    RFP dumped into a 6-touch drip.

    Brand risk on a 10-day close.

    auto-sent

  • Jordan Hale

    Internship ask got a sales pitch.

    Junk never filtered.

    auto-sent

  • Aiden Cole

    Crypto spam got “thanks for reaching out.”

    The bot emailed a scammer.

    auto-sent

Live mock inbox

Same fixtures. Click a row to stamp the gate. Public path stays /.

LeadCompanyScoreGatesent
Maya Chenmaya.chen@harborline.exampleHarborline Analyticshot99pendingfalse
Sam Ortizsam.ortiz@northwell.exampleNorthwell Studiohot86pendingfalse
Priya Nandakumarpriya.n@lumenfield.exampleLumen Field Opshot99pendingfalse
Theo Briggstheo@redcedar.exampleRedcedar Labshot99pendingfalse
Elena Vosselena.voss@quillcopper.exampleQuill & Copperwarm64pendingfalse
Jordan Halejordan.hale@mailinator.example—junk26pendingfalse
Aiden Colewinner@cheapcrypto.example—junk8pendingfalse

Mock ingest

Form JSON or CSV. Empty / broken payloads skip CRM and land in retries.

Bad payload next: Empty object

Retries (latest)

Bad or empty payloads never create a send. They only log.

  1. Empty JSON object

    form_json · attempt 1 · 2026-11-10 12:39:07 UTC

    {}
  2. CSV has no data rows

    csv · attempt 1 · 2026-11-10 12:39:07 UTC

    name,email,message