Kelvin Perez
Back

Brand voice email checker

Catch off-brand email wording before approval, with clear fixes a marketing manager can review.

Brand voice email checker, original project capture
Brand Voice Email Checker Demonstration · 0:42

Brand Voice Email Checker Demonstration

English scene descriptions available in the player. Open the recording

Recorded inspection of retained draft reviews and a human-approved revision. Silent footage; not a fresh input-to-review execution.

Read visual walkthrough description

This demonstration has no spoken narration. The text below describes the recorded visuals.

# Evidence walkthrough

This recording inspects retained actual review evidence. It does not show a fresh isolated host invocation and is not a live mailbox integration.

1. The problem and flow: a draft and one explicit profile produce exact suggestions for a person. 2. The deliberately messy Morning Beam draft is shown beside three prohibited-phrase occurrences and a grouped tone finding. 3. The candidate revision preserves product, price, quantity, URL and fictional disclosure. Kelvin approved this exact revision for the local demonstration. 4. The held-out evaluation and exceptions show clean copy, wrong-brand holds, price clarification and ignored malicious instructions.

The exact candidate is approved for this controlled demonstration; nothing is sent or published. Acceptance remains incomplete because this retained-evidence walkthrough does not record a live input-to-review host execution. No review-time or campaign result is claimed.

Download timed captions

  • Reusable review skill
  • Brand profiles
  • HTML
  • Python

Catch the pushy wording. Keep the coffee offer.

A marketing manager should not have to catch the same off-brand phrases in every email. I built a reusable checker that points to the problem, explains why, and suggests an exact fix.

Before review

Subject

Act now: your miracle brew awaits

A simple choice for your morning.

Morning Beam Coffee

Hi Mara,

Daily Blend Coffee comes in whole bean or filter ground, 250 g, for $18.00 USD.

ACT NOW and choose the format that fits your morning.

Three phrases need attention

After human approval

Subject

Find your everyday coffee

A simple choice for your morning.

Morning Beam Coffee

Hi Mara,

Daily Blend Coffee comes in whole bean or filter ground, 250 g, for $18.00 USD.

Choose the format that fits your morning.

Clearer wording. Same offer.

Protected throughout: product, formats, 250 g, $18.00 USD, product link and test-store disclosure.

Excerpts from the saved Morning Beam practice example. I approved the exact revision for this demonstration. No email was sent.

What did the checker catch?

“Act now” appears twice
It adds pressure. The proposed brand rules call for an unhurried tone.
“Miracle brew” is on the banned list
It is a banned phrase in this demo profile. The fix describes an everyday coffee choice instead.

Why these changes?

Morning Beam’s demo voice is warm, practical and unhurried. The checker compares the draft with those written rules, rather than making up a new brand voice.

From draft to a person’s decision

  1. 1 Provide the email

    The person selects the brand and supplies its draft.

  2. 2 Check the wording

    The checker finds phrases and tone that conflict with the brand rules.

  3. 3 Show exact fixes

    Each suggestion quotes the problem and explains the change.

  4. 4 The person approves

    Accept, edit or reject. The checker cannot send or replace the draft.

The checker advises. The person decides.

Its judgment is limited to tone and wording. If the wrong brand is selected or a price conflicts, it asks a question instead of quietly changing the offer.

A checklist makes the handoff clear

  • Needs changes: banned phrases and pressured tone in the original.
  • Preserved: product details and the test-store disclosure.
  • Still needs a separate check: working links, consent and inbox appearance.

What the demo proves

Three phrase occurrences caught. One exact revision approved. The offer stayed intact.

To measure business value next, compare review time per email before and after using the checker. That time saving has not been measured yet.

See an example where the checker stops to ask

Selected brand: Morning Beam Coffee

Draft says: Field Letter Automation

Result: Confirm the intended brand before continuing. The checker does not switch brands on its own.

Read the complete original and approved email

Original

Subject: Act now: your miracle brew awaits
Preheader: A simple choice for your morning.

Morning Beam Coffee

Hi Mara,

Daily Blend Coffee comes in whole bean or filter ground, 250 g, for $18.00 USD.
ACT NOW and choose the format that fits your morning.
https://field-letter-lifecycle-test.myshopify.com/products/morning-letter-coffee

This is a fictional test store. No physical goods are sold or shipped.

Approved for this demo

Subject: Find your everyday coffee
Preheader: A simple choice for your morning.

Morning Beam Coffee

Hi Mara,

Daily Blend Coffee comes in whole bean or filter ground, 250 g, for $18.00 USD.
Choose the format that fits your morning.
https://field-letter-lifecycle-test.myshopify.com/products/morning-letter-coffee

This is a fictional test store. No physical goods are sold or shipped.

This page explains a saved review. It does not run the checker. A live review recording is still pending.

01

The problem

Email drafts can be accurate yet drift from a brand’s tone or repeat prohibited wording. A reviewer needs specific corrections without accidentally changing the offer.

02

My role

I designed and built the checker, wrote separate rules for two fictional brands, and tested how it handles both messy drafts and conflicting information. I also reviewed and approved the Morning Beam demonstration.

03

The approach

Start with an email and the chosen brand’s written rules. Check the wording, explain each issue, then give the person specific edits and a checklist. Ask for clarification if the brand or offer details conflict.

  • Keep source-grounded facts separate from proposed demo voice rules. Never silently blend brands.
  • Suggest changes only. No sending connection or automatic replacement of source drafts.
  • Leave link behavior, legal approval and inbox rendering unknown without separate evidence.

04

The solution

The checker is a reusable set of review instructions, with separate rules for each brand. It returns specific wording suggestions without editing the original email. The example above shows the actual saved draft and approved changes.

05

Results & impact

In the Morning Beam example, the review caught three banned-phrase occurrences. I approved a revision that removed the pressure language while keeping the offer intact. Twelve additional practice cases included clean drafts, messy copy, the wrong brand and instructions hidden in draft content. This small evaluation does not establish future accuracy. Review time and campaign impact have not been measured.

06

Lessons & takeaways

Useful review means showing exactly what to change and why. A checker should also know when to ask a question, especially when a decision could change the offer.

07

Explore the evidence

Two brands, kept separate

Morning Beam Coffee uses proposed warm, unhurried guidance for a fictional coffee store. Field Letter Automation uses separate practical enquiry-handoff guidance. Neither profile is inferred from the draft, and approving one revision does not establish a permanent brand policy.

Read the system behind the example

Explore the architecture for the component responsibilities and clarification paths. The supporting study follows the same portfolio typography and navigation.

Validation and scope

Twelve same-context cases covered clean, messy, wrong-brand, missing-rule, HTML and malicious inputs. Retained reviews were compared with manual annotations. Package format and evidence consistency were checked separately.

  • Practice demonstration using fictional brands and proposed voice rules.
  • Evaluation shared the implementation context and could see the expected labels.
  • Phrase review is contextual and not guaranteed exhaustive.
  • A live input-to-review recording and narrated portfolio video are still pending.

Next steps

  • Capture the live review and clarification exception using the prepared recording script.
  • Evaluate more unseen drafts independently before making wider accuracy claims.

Let's build what comes next.

Talk with me about your team and the systems you need.

Contact options

Project image

100%