Reviewing and grading copy in the AI build

Step 8 of the AI build shows you the finished email sequence for each segment, a quality grade from the automated reviewer, and tools to request changes before anything is sent.

9 min read · Updated Jun 29, 2026

Overview

After the AI completes the Strategy step (step 7), it moves into Email copy - the final step of the build. This is where the agent writes the full email sequence for every segment and immediately runs it through an automated review. You see the output in a split-panel view: a sidebar listing each segment and email card on the left, and a Gmail-style preview of the selected email on the right.

A floating badge in the bottom-right corner of the preview shows the quality score for the active segment. You can click it to open a detailed grade breakdown, scrub through revision rounds, and leave inline comments on any paragraph. Once you are satisfied, the build concludes and the campaign is ready to configure and launch.

The Email copy step is step 8 of 8 in the build rail at the top of the page, labeled Email copy.

Before you start

You need an active AI build session at /campaigns/build. The Email copy step only appears after the agent has finished steps 1-7. Earlier steps cannot be skipped mid-session.

No special role or plan level is required beyond having access to the "Create with AI" feature.

How to review the email sequence

Navigate segments and emails

Open /campaigns/build and wait for the agent to reach step 8. The step rail dot labeled Email copy fills in and the split-panel appears.
In the left sidebar, a dropdown at the top lets you switch between segments. Select the segment you want to review.
Below the segment selector, individual email cards appear in order - Email 1, Email 2, and so on - with the wait time between them shown between cards. Click any card to load that email in the preview on the right.
If a step has A/B variants, version tabs appear on the card (A, B, etc.). Click a tab to switch the preview to that variant.
Use the lead switcher at the top of the preview panel to see the email rendered with a specific lead's personalized lines filled in.

Read the grade badge and open the grade panel

Look at the bottom-right corner of the preview panel. A dark badge shows a number such as 4.7 - this is the overall quality score for the active segment out of 5.
While the agent is still processing a segment, the badge shows a spinner with a label such as Round 1 - writing or Round 1 - reviewing. Wait for it to settle before acting.
Click the badge to open the grade panel. The panel shows:
- The overall score and, if the copy improved across rounds, a note showing the starting score.
- Round tabs labeled R1, R2, and Final (only rounds that ran appear). Click a tab to see the emails and scores from that round in the preview.
- Five category bars: Clarity & Flow, Brevity, CTA Strength, Relevance & Proof, and Deliverability. Each bar shows the score for that category in the currently selected round.
Click the badge again to close the panel.

Read inline AI comments

The AI reviewer attaches comments to specific paragraphs. When an AI comment exists on a paragraph, an indicator appears in the preview. Hover over the paragraph to see the comment text and its severity level.

Add your own comments

Hover over any paragraph in the preview. A comment icon appears at the edge.
Click the icon and type your note. Press Enter or click outside to save it.
Your comment appears inline alongside any AI comments on the same paragraph. Comments are anchored to the active round.

Request a revision

After the agent finishes reviewing a segment (the badge shows a solid score, not a spinner) and you have added at least one comment of your own:

A button labeled Revise from my notes appears in the bottom-left corner of the preview.
Click Revise from my notes. The chat pane sends your collected comments to Ken as a revision request for the active segment.
The badge switches to a blue spinner labeled Revising from your notes. A new round is created. When scoring completes, the Final tab updates and the score delta ("up from X") reflects the improvement.

If you prefer to write a free-form revision request, type directly in the chat pane instead and describe the changes you want.

How it works

The Email copy step runs a two-AI pipeline per segment. One AI agent writes the email sequence based on the segment strategy, ICP, and offer. A separate reviewer AI then scores the draft against a five-category rubric and a fixed checklist of hard rules.

The five grade categories are:

Clarity & Flow - does the copy read naturally and transition well between emails?
Brevity - is the static body within the word limit (target under 80 words per email, under 55 for follow-ups)?
CTA Strength - does every email close with a direct ask rather than a diagnostic question?
Relevance & Proof - does the copy speak to the specific segment's pain points and offer evidence?
Deliverability - does the copy avoid spam-trigger words, banned phrases, and problematic formatting?

The reviewer assigns a score from 1 to 5 on each category and computes an overall score. A score of 4.5 or higher is required to pass. However, a high score does not override hard rule violations. If any hard rule is broken - such as using a banned phrase, an incorrect variable format, or a hand-typed signature instead of the system token - the round fails regardless of the number.

When a round fails, the reviewer emits structured revision notes per email. The writer agent reads those notes and produces a new draft. The loop runs up to three times automatically. The grade panel shows each round as a separate tab (R1, R2, Final) so you can compare how the copy changed with each pass.

Human revision notes (from the Revise from my notes button or the chat) create an additional round using the same loop. The badge labels these rounds differently - a blue spinner says Revising from your notes - so you can tell automated rounds from human-triggered ones.

The personalized first line (the AI-variable opener) is written and evaluated separately. The reviewer checks that the static body's opening sentence reads as the natural continuation of the personalized line, not an abrupt topic change.

Troubleshooting & debugging

The badge is spinning and has not settled after several minutes.

Likely cause: the reviewer is still running or the AI provider is slow. The loop can run up to three automatic rounds.
What to check: stay on the Email copy step. The badge text shows the current state - "Round 1 - writing," "Round 1 - reviewing," etc. Each round takes up to a minute or two. If the badge has shown no change for more than five minutes, reload the page and return to the build. The session resumes from the last saved state.

The badge shows a score but the grade panel only has one tab (no R1/R2 history).

Likely cause: the copy passed on the first round, so only the Final tab exists.
This is expected behavior. A single tab means no revisions were needed.

The "Revise from my notes" button does not appear after I add a comment.

Likely cause: the segment review has not settled yet (the badge is still spinning), or no human comments exist for the active segment and round.
What to check: confirm the badge shows a static score, not a spinner. Confirm you saved at least one paragraph comment for the currently selected segment. Comments from a different segment do not trigger the button for the active segment.

The score is 4.5 or higher but the round is labeled as failed.

Likely cause: a hard rule violation (banned word, wrong variable format, hand-typed signature).
What to check: open the grade panel and read the AI comments on the draft. Hard violations appear as "reject" severity comments. The body text of the comment names the specific rule that was broken. Address those in your revision request.

After clicking "Revise from my notes," the score went down.

Likely cause: the revision instruction changed parts of the copy that were previously strong, or introduced a new rule violation.
What to check: scrub between rounds in the grade panel to compare category scores. The round tabs show the score for that round. Use the chat pane to clarify which specific part to change, rather than sending broad revision notes.

A segment shows a score in the sidebar but the grade panel says no review data.

Likely cause: you switched segments mid-revision and are viewing a segment that has not finished its first round yet.
What to check: the sidebar segment score updates live. If a segment's score is missing (shown as a dash or loading indicator), wait for that segment's round to complete before opening its grade panel.

The preview shows placeholder text like {{First Line}} instead of a real personalized line.

This is expected. Personalized AI-variable lines are generated per-lead at send time, not during the build. Use the lead switcher at the top of the preview to see the email rendered with sample lead data pulled from your ICP.

FAQ

Can I approve individual emails or variants?

Yes. Each email card in the sidebar has a checkmark you can toggle per variant. Approval state is local to your session and is informational - it does not gate the build from proceeding.

Does leaving a comment send anything to the campaign right away?

No. Comments are stored locally in the build session. They only act on the campaign copy when you click Revise from my notes or paste them into the chat.

How many revision rounds can the system run automatically?

Up to three automatic rounds run without any input from you. Human-triggered revisions (via Revise from my notes or the chat) each add one round on top, with no hard ceiling.

What happens if the copy never reaches 4.5 after three automatic rounds?

The loop stops and the best scoring draft is kept as the Final round. The badge shows the final score even if it is below 4.5. You can still review it, add comments, and trigger a human revision to push it further.

Can I jump back to an earlier round's copy after the final round is done?

Yes. Click any round tab in the grade panel to render that round's emails in the preview. Selecting an earlier round does not change what is saved - it is a read-only view for comparison.

Is the copy written per-segment or for the whole campaign at once?

Per segment. The AI runs the write-review loop separately for each segment, in parallel. That is why the grade badge and panel are scoped to the currently selected segment in the sidebar.