Email marketing is a volume game played in words. Every week there is a campaign to write, a subject line to bet on, a welcome flow to keep fresh, a lapsing segment to win back, and a set of numbers to read once everything sends. The writing is repetitive, the math is fiddly, and the part that matters most, the subject line, is six words you are guessing at under time pressure.
An AI agent fits this shape of work. The difference from a plain chatbot is that this one remembers your brand voice and your offer so you stop re-pasting them, searches the live web for the seasonal hook or the competitor send worth reacting to, and reads your CSV export with real Python to calculate open and click rates instead of eyeballing them. The Keimodel angle is sharpest exactly where it counts: a subject line is the highest-leverage thing you write all week, so instead of trusting one model's taste, you can have several models each pitch lines for the same email and pick the winner on merit.
Set your voice, your list, and your offer in Memory once, then the prompts below cover the campaign, the flow, the subject-line test, and the monthly read in a few minutes each. You own the calendar and the send button. The agent drafts, counts, and never ships without you.
Capabilities this leans on
Do this first. Every subject line, sequence, and segment gets sharper once the agent knows your brand, your list, and your offer.
Remember these facts about our email program: we're [Brand], a [what you sell] brand emailing from [domain] through Klaviyo. Brand voice is warm, plain, and a little dry, never hypey, no exclamation marks, prices written like $48. Our list is about 22,000 subscribers, average open rate runs near 41% and click rate near 1.8%. Main segments are engaged (opened in the last 30 days), lapsing (60 to 90 days), and dormant (90 days plus). Our standard new-subscriber offer is 10% off the first order, code WELCOME10. Always give me a plain-text variant alongside any designed email, and never invent a stat or a claim we can't back up.
One announcement becomes the email, a plain-text variant, and a reusable template.
Write a campaign email announcing our fall stoneware restock, landing Thursday. In our voice: a subject line and a preview line, a 120-word body with one clear CTA to shop the collection, and a plain-text variant. The hook is 'back in the three colors that sold out in spring'.
Give me five subject line options for it, a mix of curiosity, plain-spoken, and benefit-led, each under 45 characters, and tell me which two you'd A/B test and why.
Save this as a Skill called 'Campaign brief to email' that takes a one-line announcement and returns the subject, preview, body, plain-text variant, and five subject lines.
What you get: A send-ready campaign in both designed and plain-text form, plus a saved Skill that turns the next announcement into a draft in one paste.
Map the whole nurture flow, then write every email in it.
Design a 4-email welcome sequence for new subscribers who took WELCOME10 but haven't bought yet: email 1 the brand story, email 2 bestsellers, email 3 social proof, email 4 a soft nudge before the code expires. For each, give me the timing, the goal, a subject line, and the full body in our voice.
Write a shorter 3-email version for subscribers who already bought once, focused on the second purchase, with no discount.
For email 4, write two versions, one that leads with urgency on the expiring code and one that leads with the product, so I can test which converts.
What you get: A complete welcome flow with timing and copy for each email, plus a tested variant on the email that drives the first sale.
The subject line is the six words that decide the open, so let several models compete for it.
Here's the body of Thursday's restock email. Pitch me eight subject lines under 45 characters, no emoji, no clickbait, in our voice. [paste body]
Now run that same prompt through two other models and put all the subject lines in one shuffled list so I can pick the strongest three on merit, then reveal which model wrote each.
Rank my chosen three for likely open rate, explain the reasoning, and tell me which to set as the A/B test winner.
What you get: A blind shortlist of subject lines from multiple models, judged on the line itself rather than on which model you trust.
Win back the dormant segment before you have to scrub it.
Write a 3-email win-back flow for subscribers who haven't opened in 90 days: email 1 'are we still a fit', email 2 a genuine best-of with one offer, email 3 a last call before we stop emailing them. Keep each under 90 words, in our voice.
Write an honest, not gimmicky, subject line and preview line for the last email, plus a one-line resubscribe CTA.
Draft the sunset note that goes to anyone who still doesn't open after the flow, confirming we've paused their emails and how to opt back in.
What you get: A full win-back flow and a clean sunset note, so a stale segment either re-engages or comes off the list on good terms.
Same news, the right framing for each part of the list.
Take Thursday's restock email and rewrite the opening and the CTA for three segments: first-time buyers, repeat customers, and VIPs who've spent over $500. Keep the core news the same, change the framing, and lay them out side by side.
Write the segmentation logic in plain English I can hand to whoever builds the segments in Klaviyo: who's included and who's excluded in each.
What you get: Three on-brand versions of one campaign and the plain-English rules to build the segments behind them.
Drop in the export and get the read your boss asks for, with the math done right.
I've uploaded last month's campaign export CSV from Klaviyo. For each send, calculate open rate, click rate, and click-to-open rate, then tell me the three best and three worst by click-to-open and what the winners have in common.
From that same file, chart click rate over the month and flag any day-of-week or send-time pattern in the opens.
Write a 150-word performance summary for our monthly marketing recap: what worked, what didn't, and the one change I'd make next month, grounded in the numbers.
What you get: Correct rates for every send, a chart, and a recap you can paste straight into the monthly report.
A planned calendar waiting every Monday, drafted from what's coming up.
Every Monday at 8am, search for any notable retail or seasonal moments in the next two weeks and draft a send plan: two campaign ideas with angles and subject lines, plus a reminder of which automated flows are currently live.
Take idea one from this week's plan and draft the full campaign using my 'Campaign brief to email' Skill.
What you get: A fresh, timely send plan in your inbox every Monday, and a one-step path from any idea on it to a finished draft.
Content marketing
Draft long-form, atomize it everywhere, and compare models to pick the best line.
E-commerce / DTC owners
Pricing math, product copy in your voice, competitor research, and customer replies.
Marketing agencies
Briefs, calendars, ad variations, and client reporting at agency speed.
Open the Agent, paste any prompt above, and change the details to fit your business.