We sent 275 cold emails with an AI agent. The telemetry is humbling.

Description and use

The 275‑email cold outreach experiment using an AI agent achieved a 91.8 % delivery rate but zero human replies, revealing a copy issue rather than automation failure.

Problem

Achieving measurable cold‑email outreach while ensuring compliance, accurate telemetry, and privacy‑respectful tracking.

Outcome

The experiment achieved a 91.8 % delivery rate, but zero human replies, highlighting a product‑and‑copy issue rather than automation failure. The pipeline demonstrated accurate telemetry and compliance enforcement.

Workflow

• 1. Vet business domains and compile one role address per business, ensuring no free‑mail or scraped personal emails and maintaining a suppression list.
• 2. Enforce a hard send‑cap in the sending script (per‑day and per‑month ceilings).
• 3. Send 275 emails, limiting to two touches per contact (currently all first touches).
• 4. Record send ledger rows and exclude internal pipeline self‑tests.
• 5. Verify delivery by joining provider events API with send ledger via message‑id, ensuring only terminal events are counted.
• 6. Preflight lint of body characters to avoid encoding regressions (e.g., em‑dash causing 550 errors).
• 7. Suppress confirmed hard bounces to prevent retries.
• 8. Implement a self‑hosted, aggregate‑only tracking pixel (day + path counts, no per‑visitor storage) for open‑rate measurement.
• 9. Monitor telemetry: delivered, permanent failures, hard bounces, replies, opt‑outs.
• 10. Apply kill‑gate on 2026‑11‑04 if replies remain below threshold, then re‑point capacity.

Evidence

275 countable sends to business domains we researched and vetted by public criteria (no scraped personal emails, no free‑mail, one role address per business, a real suppression list).

Source

delivered (receipt confirmed) 167 — 91.8% (167/182)

Source

replies from real humans 0

Source

We wanted open rates. Our API keys could not enable the provider's open‑tracking setting... we shipped a self‑hosted, aggregate‑only tracking pixel

Source