A single test email can verify layout and links. It cannot verify a campaign where the subject or body changes for every recipient. Recipient-specific generation needs recipient-specific review evidence and a clear definition of done.
The goal is not to make a person reread identical fixed copy hundreds of times. It is to focus attention on the parts that vary, automatically flag predictable defects, and preserve an exact snapshot of what will be sent.
Review the complete recipient email
Never approve only the generated passage. Read the subject, greeting, fixed message, personalized passage, CTA, signature, footer, and links as one email. Personalization that looks good in isolation can still repeat a fact or create an abrupt transition.
The preview should use the actual recipient row and the same campaign version that sending will use. A generic sample is useful while writing, but it is not final approval.
Use a five-part review rubric
Use the same rubric for manual edits. A draft that was approved and then changed should return to review because the approved artifact no longer exists.
- Accuracy: every factual statement is supported by the available contact data or approved campaign information.
- Relevance: the adapted passage explains why this message matters to this recipient without overreaching.
- Consistency: the offer, claims, CTA, voice, and fixed message remain aligned across recipients.
- Completeness: variables, greeting, links, signoff, unsubscribe content, and required address information are present.
- Permission: the recipient is eligible for this campaign and is not unsubscribed, bounced, complained, or otherwise suppressed.
Automate the predictable checks
Quality checks should catch unresolved merge fields, empty AI passages, missing CTAs, missing greetings, suspicious length, and required footer problems before a reviewer opens the draft.
These checks are gates, not scores to optimize. A short email is not automatically good, and an email that resolves every variable may still make an unsupported claim.
Prioritize where human attention goes
Filter drafts that need review first. Prioritize missing context, unusual generated length, sensitive lifecycle stages, high-value accounts, first-time campaign instructions, and drafts changed after approval.
For larger campaigns, review every flagged draft and a representative sample of ready drafts across segments. Increase sampling when the audience, data source, instruction, or offer is new.
Use states that describe real decisions
Generated, Ready, Needs review, and Done should mean different things. Generated describes an artifact. Ready means automated checks passed. Needs review means a defect or decision remains. Done means a person accepted the current revision.
Reopen a draft when it changes. Store the reviewer and time for consequential campaigns. Block scheduling when the active campaign version has missing or unresolved previews.
Generated -> automated checks
Checks pass -> Ready
Check fails or edit occurs -> Needs review
Reviewer accepts current revision -> Done
Template changes -> previews become staleFinish with campaign-level readiness
Commercial email also needs accurate routing information and subject lines, an appropriate opt-out method, and prompt suppression of opt-outs. The FTC notes that CAN-SPAM applies to business-to-business commercial messages as well as consumer messages.
- The intended contact list and campaign version are attached.
- Every eligible recipient has a saved preview.
- No preview contains an unresolved blocking issue.
- Required human approvals are complete.
- The sender domain and From address are verified.
- Consent, unsubscribe, bounce, and complaint checks pass.
- The scheduled volume fits the plan and warm-up limit.
- The final send job snapshots the approved content.