Skip to content
mailprompt
Start free trial
FeatureUpdated

A human approves every marketing send

Mailprompt separates drafting from sending. The chat, the API and any connected client can create a draft, but a marketing send only leaves the account after a person approves the current version. Approvers see the audience, the rendered email, the subject line and the schedule, and a later edit needs a new approval.

Key facts

  • The CAN-SPAM Act gives recipients the right to stop commercial email and requires senders to honor opt-out requests. Source
  • Google's bulk sender guidelines require a one-click unsubscribe mechanism for marketing messages sent at volume. Source
  • Yahoo's sender best practices also ask bulk senders to authenticate their mail and make unsubscribing easy. Source

What this page covers

This page is part of the Mailprompt features set and explains the approval gate that sits between a draft and a send. It is the least glamorous part of the product and the one most likely to keep a store out of trouble.

Drafting is cheap; sending is not

A model can write a welcome email in seconds. It can also write a subject line that sounds like a final notice, a discount that erodes the margin or a promise the warehouse cannot keep. The cost of catching those problems before a send is a few minutes of review. The cost of catching them afterwards is a support queue, an unsubscribe spike and a deliverability problem that outlives the campaign.

Mailprompt is built around that asymmetry. The drafting tools are fast and available to the chat, the API and any connected client. The send gate is narrow and human. Speed belongs on the creative side; caution belongs on the side that reaches the customer.

What the approver sees

An approval screen that shows only a subject line is not a review. The approver needs enough context to answer three questions: who receives this, what does it say, and when does it go out.

Mailprompt presents the audience with its size and its defining conditions, the rendered email and its plain-text version, the subject line and preview text, the send schedule and the flow or campaign the message belongs to. If the draft includes a coupon, the approver sees the coupon rule rather than a code copied into the copy. If the draft depends on a product feed, the approver sees which products were selected.

The view also shows what changed since the last approved version. A subject-line edit is a small diff. A changed audience is not, even if the email body is identical. Treating those two changes as equivalent is how a routine approval turns into an accidental send to the whole list.

The rule has to hold everywhere

A review step that only applies to the web interface is not a rule. If the API can send directly, a script becomes the bypass. If a connected client can call a send tool without approval, the chat becomes the bypass. If an internal job can mark a draft as approved, the queue becomes decoration.

Mailprompt applies the same gate to every path. The web app requests approval. The API exposes a draft and an approval resource, not a hidden send shortcut. A connected client can create and revise drafts but cannot record the human approval. The send worker checks the approval record before it queues a marketing message, so a direct database write is not enough to make a send legal.

The failure mode is deliberate: if the approval record is missing or stale, the send fails. It does not fall back to “send anyway” or “send a test to the owner”. A blocked send is visible and fixable; a silent bypass is not.

Rejecting, revising and re-approving

Approval is not the only outcome. An approver can reject a draft with a note, ask for a narrower audience, request a different subject line or send the draft back to the chat for another pass. The note becomes part of the draft history, so the next reviewer does not have to reconstruct the conversation.

A revised draft is a new version. The previous approval does not carry over, even when the change looks small. This is stricter than some teams expect, but it keeps the rule simple: the thing that was approved is the thing that sends. If the content changed, the approval has to change with it.

For a recurring campaign, each scheduled send still needs the approval that matches its version. A schedule is not a blanket approval; it only decides when an approved send may go out. If the template or the audience changes, the approval no longer covers the new version, and the campaign returns to review.

What approval does not fix

A human reviewer is not a deliverability guarantee. The message still needs authenticated domains, a clean list and a working unsubscribe link. The approval gate reduces the chance of an obvious mistake; it does not replace the send gates for suppression, consent, frequency caps or domain verification.

Approval also does not make a claim true. If the copy says a product ships in two days and the warehouse cannot meet that promise, the approver has to catch it. The product can show the source of a product claim, but it cannot verify the warehouse's capacity. That is a human judgement, which is why the review step exists.

How this connects to the rest of the product

The approval queue is where the chat-built flows meet the send path. It is also where the revenue reporting gets its meaning: a send that was approved, scheduled and delivered can be compared with the orders that followed it. Without a recorded approval, the result is hard to trust and impossible to audit.

The domain setup work sits earlier in the chain. An approved message still fails if the sending domain is not authenticated. Approval, authentication and attribution are three different controls; none of them substitutes for the others.

Frequently asked questions

Who can approve a send?

A person with the OWNER or ADMIN role on the account. The draft records who approved it and when, so the answer is available later. Other roles can draft and comment but cannot record the approval. If the account has more than one owner or admin, any one of them can approve; the product does not require a second signature.

Can the chat approve its own draft?

No. The chat can create and revise a draft, explain it and ask for changes, but it cannot record a human approval. The approval action is tied to a signed-in OWNER or ADMIN. This is deliberate: a model that can both write and approve has no meaningful review step, even if it usually makes good choices.

What happens if someone edits an approved campaign?

The edit creates a new version and the approval is no longer valid for that version. The campaign returns to the review state until a person approves the change. The same rule applies to a change in audience, subject line, send time or email content, because each of those can change who receives the message and what it says.

Does the approval rule apply to the API and other clients?

Yes. A client can create a draft and request approval, but it cannot send without one. The public API exposes the draft and approval workflow rather than a bypass. If a client tries to send an unapproved campaign, the request fails with an approval error instead of falling back to a direct send.

What happens if no one approves before the scheduled time?

The send waits. There is no auto-approval timer that turns a draft into a send after a delay. If the scheduled time passes, the campaign stays unapproved until a person approves it or the team reschedules it. This is intentional: an approval that silently expires into a send is not an approval, and a deadline should not be able to bypass the rule.

Sources