Writing and scheduling marketing emails with AI

How the bot works

Nine steps. Gathering the approved material, building the segment, running the claims and disclosure checks and holding the cadence are ordinary code and cost almost nothing to run; the model is only asked to turn material you already stand behind into an email somebody will read.

1. It starts from material you have already approved. Your own pages, your posts, the product records, the event that is on the calendar, the offer somebody signed off. A topic is not a source, and neither is a competitor's email.

2. The send has a reason before it has a subject line. The newsletter slot that comes round, the product that went live, the date that is approaching, the change you published. If nothing in your own material is new, the bot reports an empty slot rather than writing something to fill it.

3. Every factual sentence names where it came from. This line is that product field, that page, that event record. A reviewer can check a claim without opening anything else, and anything with no source behind it is not written at all.

4. Your banned claims and your disclosures run before anybody reads it. The phrases you have taken out before, the comparisons you do not make, the disclosure a category has to carry, the footer and unsubscribe block you configured. A draft that trips one of these arrives flagged with the rule named.

5. Subject lines come as options. Three or four, with the preview text, and a note on what each one leads with. Choosing is a person's job, because the subject line is the part of the email that makes a promise on your behalf.

6. The segment comes out of your own list. Who gets it is a rule against data you already hold — customers of a product, people who booked, the region, the people who asked for this. It adds no address, it appends nothing from outside, and an address that opted out is not in any segment.

7. The cadence is a limit it cannot move. How often this list hears from you, the quiet hours, the blackout dates, the cap per person per week. The limit lives in your configuration rather than in a prompt, so a model cannot talk itself past it.

8. A named person approves the exact send. The draft waits with its sources, its flags, the chosen segment and the schedule. Somebody approves the version they read, and that version is what goes out, at the time it was approved for. An unapproved draft is never sent.

9. It tells you what it did not send. The slot it left empty, the claims it would not write, the segment it skipped because consent was unclear, the product with nothing new to say. That list is the useful output in the first month, because it is the list of sends nobody could put a source behind.

What you configure

It can be configured to run in any business. This is what that actually means — the list of things that are yours, not ours.

Your sources of material

Which pages, posts, product records, calendars and documents it may draft from — and which are out of bounds entirely.

Your claims and disclosures

The phrases you never use, the comparisons you refuse to make, the disclosure a product or a category must carry, and the footer every email ends on.

Your lists and segments

Which lists exist, what each one consented to, the rules that build a segment, and the fields that prove consent. Nothing outside your own data builds an audience.

Your cadence and quiet hours

How often each list hears from you, the per-person cap, the send windows, the blackout dates, and what happens when two campaigns want the same week.

Your voice

The words you use for your own products, the ones you refuse, how long an email runs, and whether it reads as a letter or a list.

Your templates

The layout, the blocks, the image rules, the link conventions, and what the mobile version has to look like.

Your subject-line rules

How many options, what they may promise, the words you keep out, and whether emoji are allowed in your house style.

Your approvers

Who may approve a send, who has to see a campaign that touches a sensitive category, and which lists always need a named person.

Your release switches

Which campaign types may be drafted automatically, which wait for somebody to ask, and whether sending is on at all. Sending starts off.

Transparent, and configurable by you

The dark band below is what the system will not do. This is what you can see and change — and it is how we build everything, not something invented for this page.

You see what it did, and why

Every factual line in a draft names its source, every flag names the rule that fired, and every send records the approved version, the segment it went to, the schedule and who approved it. A campaign can be audited against your own material rather than taken on trust.

Versioned prompts you can tune yourself

The claims list, the disclosure rules, the segment definitions, the cadence limits and the prompts behind the drafting are yours to tune. A change is a reviewable diff, replayed against campaigns you have already sent so you can see what it would have written differently, and rolled back the same way.

A human in the loop wherever judgment actually matters

Gathering the material, drafting, checking the claims, building the segment and holding the schedule are automatic. Deciding what to say, which subject line promises what, who should hear from you and whether this goes out at all are a person's — that boundary is a property of the system, not a promise about how carefully we'll use it.

Every run logged, along with what it cost

One row per campaign: the material it read, every line with its source, the flags it raised, the segment and the count it went to, the approver, the time it was sent and what it refused to write — plus token usage and cost for that run.

It can be configured to run in any business. The material and the rules are yours: your pages, your product data, your claims list, your lists, your cadence, your approvers. The campaigns on this page are one configuration of it, not the limit of it.

Works with your tools

This drafts into the email platform you already pay for. We're not selling you a sending platform, and your list stays where it is.

• Mailchimp
• Klaviyo
• HubSpot
• Constant Contact
• Your website or CMS
• A product catalog
• A booking or events calendar
• A spreadsheet of subscribers

Works well for

• Ecommerce stores — New arrivals, back in stock and the weekly send all pull from product data that already exists.
• Restaurants — A seasonal menu, an event or a closure needs the same list told, often on short notice.
• Real estate agencies — Listings and open houses are already published, and the email is those facts for a different audience.
• Nonprofits — The newsletter, the appeal and the event invite all come out of program work that is already written up.
• Dental practices — Appointment reminders run through a different system, so the newsletter is the one that waits.

How a rollout works

Nothing is sent on day one. The first milestone is a month of campaigns you have already sent, drafted again from your own material, so you can compare them line by line with what actually went out.

1. 01Discovery

   The quote call, then a working session. We go through the material it may draft from, your claims and disclosure rules, the lists and what each one consented to, your cadence and quiet hours, your templates, who approves, and what must never be drafted automatically. You get the source map, the rules and a written scope.

2. 02Dry run

   A month you have already sent, drafted again. Nothing is scheduled and nothing reaches a subscriber. You compare the drafts with what went out and read the refusal list — and the refusal list is where the work is, because the claims nobody can source are invisible until something goes looking for them.

3. 03Assisted mode

   Live on one list. The bot drafts into your platform, your marketing person edits and approves, and nothing reaches a subscriber unread. Who decides what the business says does not change.

4. 04Optional limited automation

   Optional, and narrower than it sounds. A short list of campaign types you name — the weekly newsletter from your own posts, a back-in-stock note — drafted automatically on your schedule, still unsent until somebody approves, with every hard stop in force and the kill switch. Plenty of teams stop at step 03, and that's a good place to stop.

What we don't claim

• No numbers we haven't earned. No open rate, no click rate, no deliverability figure, no revenue-per-email claim, no emails-per-hour number. We have not mailed your list. What a campaign costs to produce comes out of the dry run, and that will be your number.
• No customer claims on this page. No names, no case studies, no logos.
• It cannot make a list want to hear from you. A number measured on somebody else's accounts would not have been about yours. The bot drafts from what you already have; whether there is something there worth sending is a judgment it does not make.
• It is not your email platform and not your deliverability consultant. Your platform stays the sender and the record. Authentication, domain reputation and whether your mail lands in a promotions tab are not things this bot controls or claims to.
• Whether an email is welcome is a real argument, and not one a vendor should settle for you. Frequency, tone, how often an offer repeats: the dry run writes out every send before any of them goes anywhere, which is the easiest moment to decide a cadence is one email too long.

Questions people ask

Can AI write our marketing emails?

Yes, from material you already own. An agent can take your own pages, posts, product records and approved offers, draft the campaign or the newsletter with the source named behind every factual sentence, run your banned-claims and disclosure checks, offer subject lines as options, and hold the whole thing for somebody to approve. What it will not do is invent a claim, an offer or a date you never published.

Where do the sentences come from?

From your own material, named line by line. A draft with a claim that has no source behind it does not get written, and a sentence a reviewer cannot trace is the thing this design exists to prevent - it is your name on the email, not ours.

Does it send without us?

No. Sending starts off, and when you turn it on an email goes out only after a named person approved that exact version, for that exact segment, at that time. There is no autopilot mode, and there is no setting that adds one.

Will it email people who never signed up?

No. It builds segments from your own list data and consent fields, it adds no address, it buys and appends nothing, and an address that opted out is in no segment. Cold outreach is a different product with a different law behind it, and it is not in this build.

How does the approval step work?

The draft waits in your platform with its sources, its flags, the segment and the proposed schedule on top. One person can approve, or you can configure two: a marketing read and a sign-off for anything touching a sensitive category. What goes out is the version that was read.

What happens when there is nothing new to say?

It reports an empty slot. An email with nothing behind it is one you would not have chosen to send, so the bot would rather tell you the newsletter has no material this week than write around the gap.

Can it do the subject lines?

It drafts a few options with the preview text and says what each one leads with. It does not pick one, and it does not run a split test it decided on - the subject line makes a promise in your voice, so a person chooses it.

Is this the same as answering emails from customers?

No, and they are deliberately separate. This one writes outbound campaigns and reads no mailbox. Triaging what arrives, drafting replies and routing the awkward ones is a different bot with its own page and its own hard stops.

Is there a record of what was sent?

Yes, per campaign: the material it drafted from, every line with its source, the flags it raised, the segment and count, the approver, the schedule, and what it refused to write - plus the token usage and cost for that run.