---
name: al-596-make-ai-written-marketing-emails-consistently-good-have-t
description: "A Nov 3, 2026 AIPOST article describes a print-on-demand t-shirt brand (its Shopify dashboards show about $1.83M in 2025 sales) that runs four separate AI agents for support, email, design and ads. Its email agent is trained on a folder of the brand's top-performing past emails and a rubric (on-brand message, reuse of offers already tested). To fix inconsistent AI emails, the brand uses 'evals': the agent drafts a batch of 10, each gets a letter grade from A to D with feedback, Claude reads the grades and adjusts, and rounds repeat until outputs consistently hit A. Every later prompt change is benchmarked against the previous version before it goes live. Suggested campaigns then sit on a Suggested / Approved / Sent board and only go out after the owner clicks Approve. The brand does not name itself in the article, and the sales figures are one business's results. Use when the user wants help with graded email batches until they hit a (marketing) for a small business."
---

# Make AI-written marketing emails consistently good: have the AI draft 10 at a time, grade each A to D against your best past emails, feed the grades back and repeat until every draft scores an A, then re-test whenever you change the prompt

A Nov 3, 2026 AIPOST article describes a print-on-demand t-shirt brand (its Shopify dashboards show about $1.83M in 2025 sales) that runs four separate AI agents for support, email, design and ads. Its email agent is trained on a folder of the brand's top-performing past emails and a rubric (on-brand message, reuse of offers already tested). To fix inconsistent AI emails, the brand uses 'evals': the agent drafts a batch of 10, each gets a letter grade from A to D with feedback, Claude reads the grades and adjusts, and rounds repeat until outputs consistently hit A. Every later prompt change is benchmarked against the previous version before it goes live. Suggested campaigns then sit on a Suggested / Approved / Sent board and only go out after the owner clicks Approve. The brand does not name itself in the article, and the sales figures are one business's results.

## Inputs to collect first
Ask the user for anything below that they haven't already given you. Don't guess missing facts.

- Campaign, e.g. new listings this week / spring promo
- Audience

## Instructions

Do the task below with the user's inputs in place of each [bracket].

1. Put your 10-20 best-performing past emails (subject line, body, results) in one folder, plus a one-page rubric you write yourself: what an A email does (on-brand, one clear offer you've already tested, short) and what makes a D.
2. Ask Claude or ChatGPT: "Using these example emails and this rubric, draft 10 emails for [campaign, e.g. new listings this week / spring promo] for [audience]."
3. Grade each draft A to D yourself (or have a second chat grade them with the rubric and you spot-check), with one line on why.
4. Paste the grades back: "Here are the grades and reasons. Update your approach and draft 10 more." Repeat until most drafts are A.
5. Save the final instructions as your email prompt. Whenever you change even one sentence of it, re-run a batch of 10 and compare the grades with the last version before using it.
6. Nothing sends automatically: keep a simple Suggested / Approved / Sent list and approve each campaign yourself.

## Safety checks (share these with the user)

- Check every figure, price and total against your own records before anything is sent, posted or booked. A person approves refunds, discounts, payments and price changes.
- Read and edit the draft before it reaches a customer or goes public. If you later automate this prompt, start with "draft and notify me", not "send automatically".
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
- Say clearly which parts need a person to review or approve before anything is sent, paid, posted or filed.

## About this skill

- Works with: Claude or ChatGPT, your email platform, a spreadsheet or board for approvals
- Based on this real-world example: https://aipost.kr/posts/2026-11-03-tshirt-brand-four-ai-agents-operations/
- Free prompt from AgentPlaybook Real Estate: https://agentic-library.netlify.app/prompt/al-596-make-ai-written-marketing-emails-consistently-good-have-the/
- To install in an agent that supports Agent Skills, save this file as SKILL.md in a folder with the same name as `name` above (for example in your skills directory). Install steps for Claude.ai, Claude Code, Cursor and Codex: https://agentic-library.netlify.app/skills/