Marketing & Growth

How to launch an email marketing campaign with AI

Marketing & Growth3 AI tools6 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

Launching an email marketing campaign means stitching together audience segmentation, copywriting, subject line testing, scheduling, and post-send analysis — often across tools that don't talk to each other. For most operators, the bottleneck isn't ideas; it's the hours spent writing sequences, cleaning lists, and trying to figure out what actually drove opens versus what was just a fluke.

Email copy, in particular, feels like exactly the kind of work AI should handle. The inputs are structured — audience, offer, tone, goal — and the output is text. That makes it a natural fit for a first prompt. Beyond drafts, AI can help you map out a full sequence logic, write subject line variants for A/B tests, and pressure-test your messaging against a specific persona before you send to a real list.

ChatGPT, Claude, and Gemini are genuinely useful here. They'll write a five-email nurture sequence from a single brief, generate ten subject line options in thirty seconds, and give you reasonable feedback on whether your CTA is buried. The constraint is everything that surrounds the text: your actual contact list, your platform, your performance data. Those live outside the model and require manual handling on your end.

Skip the manual work → See the Starch version

Marketing & Growth3 AI tools6 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ChatGPTClaudeGemini

Step-by-step

1 Start with a campaign brief: paste into Claude or ChatGPT your product description, target audience (job title, pain point, where they are in the funnel), campaign goal, and desired sequence length. Ask it to output a structured sequence map — email 1 subject, goal, CTA; email 2 subject, goal, CTA; and so on.

2 Once you have the sequence map, prompt the model to write each email in full. Give it the context from the map plus your brand voice. Do one email at a time to keep the output focused; long multi-email prompts tend to produce generic filler by email three.

3 Generate subject line variants for each email. Ask for 6-8 options per email, ranging from direct to curiosity-driven to personal. Paste your best candidates back in and ask the model to predict open rates by tone — not scientific, but useful for gut-checking.

4 Paste your draft email into the model and ask it to critique it as the specific recipient. Prompt: 'You are a [job title] who receives 80 cold emails a week. What's your reaction to this email? What would make you click?' Use the feedback to tighten the copy.

5 Export your contact list from your email platform (Mailchimp, Klaviyo, ActiveCampaign, etc.) and use the model to help you write segment descriptions or tag logic — for example, 'here are my subscriber fields, suggest how to split them into three audience buckets for this campaign.'

6 After the campaign runs, export your performance data (opens, clicks, unsubscribes) as a CSV, paste it into the chat, and ask for a plain-English summary of what worked and what to change in the next send.

Prompts you can copy

Write a 4-email welcome sequence for a B2B SaaS product that helps small restaurants manage inventory. Audience: owner-operators with 1-3 locations. Goal: get them to book a demo by email 4.

Generate 8 subject line variants for an email announcing a limited-time discount on an annual plan. Audience: free users who have been active for 30+ days but haven't upgraded. Tone: conversational, not salesy.

I'm writing a re-engagement email to subscribers who haven't opened anything in 90 days. Here's my draft: [paste draft]. Rewrite it to be shorter, lead with the reader's problem, and end with a single clear question instead of a CTA button.

Here is a CSV of my last email campaign's performance by subject line: [paste data]. Summarize what the open rate and click rate differences tell me about what messaging is resonating with this audience.

Suggest a segmentation strategy for an e-commerce email list of 8,000 subscribers. I have data on: purchase history, days since last order, total spend, and product category preferences. What are the three most useful segments to target separately?

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

Your contact list and subscriber data live in Mailchimp or Klaviyo — you have to export it, paste it in, and repeat that process every time you want a fresh analysis.

Nothing persists between sessions. The sequence map you built last week, the voice notes you gave the model, the segment logic — all of it is gone the next time you open a new chat.

Post-send performance data requires a manual export-and-paste loop. You can't ask 'what's working across my last five campaigns' without assembling that data yourself first.

The model has no access to your CRM, so audience insights are generic. It can't tell you which contacts haven't heard from you in 30 days or which deals went cold after the last send.

Output consistency drifts. The tone, formatting, and structure you carefully prompted in week one won't reliably match what you get in week three without re-pasting all your instructions.

There's no connection to your actual sending platform. You still copy every email manually into Mailchimp or Klaviyo, set up the sequence logic by hand, and configure the scheduling yourself.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — for this workflow, that means an agent builds a persistent email campaign system connected to your live contact data, CRM, and analytics, so the work runs continuously instead of restarting from a blank prompt every time.

Connect Mailchimp, Klaviyo, ActiveCampaign, or ConvertKit directly from Starch's integration catalog. The agent queries your live subscriber data — segments, engagement history, tags — when it drafts or schedules, not a weeks-old export.

The Growth Analyst starter app connects to your PostHog traffic and Gmail data and emails you a weekly digest covering what's working by channel, signup trends, and where to focus next — so your campaign decisions are grounded in current numbers, not gut feel.

The CRM starter app syncs your Gmail threads and enriches contact profiles, so you can ask 'who hasn't heard from us in 30 days and fits this segment?' and get a real answer pulled from live data — not a manual filter job.

Describe your campaign sequence in plain English and the agent builds a reusable automation: 'Every time a new lead is tagged as trial-started in HubSpot, send this 4-email onboarding sequence over 14 days, then Slack me if they haven't clicked by day 10.'

Post-send analysis runs automatically. Instead of exporting CSVs and re-explaining context to the model, Starch queries your email platform's performance data live and surfaces what changed — no copy-paste loop, no lost context between sessions.

Get closed-beta access →

Toolkit