Prompt E-mail Formal

Crafting prompts for formal emails

A prompt for a formal email is the brief you give to ChatGPT or Claude so it writes business outreach, follow-ups, apologies, proposals, complaints or thank-you notes for you. The single biggest mistake is asking “write a formal email about X” and pasting whatever comes back — the result reads obviously robotic, opens with “I hope this email finds you well” and rambles for 250 words. The fix is to feed the model an explicit structure (RTCF) so it produces an email that sounds like you, not like a generic assistant.

Template that works for cold outreach, follow-up and apology alike:

Role: Senior account executive at a SaaS company.
Task: Write a polite follow-up to a prospect who missed the demo.
Context: Demo scheduled 3 days ago, no-show, no reply.
         Want to reschedule, will attach a case study.
Format: 80-120 words, subject line, single CTA,
        friendly-professional tone, no aggressive language.

Email frameworks that pair well with the prompt

Once the model knows the goal, ask it to follow a copywriting framework. AIDA — Attention, Interest, Desire, Action — is the classic sales sequence. PAS — Problem, Agitate, Solution — is sharper for pain-driven outreach. QUEST — Qualify, Understand, Educate, Solution, Take action — suits long-form. BAB — Before, After, Bridge — lands well in case studies. Specify the framework in the Format field: “Use PAS structure”.

Tone, language and audience

Formal (LinkedIn first contact, partnership proposals) — full names, no contractions, structured paragraphs. Semi-formal (follow-up, internal stakeholders) — first names, light contractions. Casual (warm intros, founder-to-founder) — first names, contractions, occasional humour.

Brazilian Portuguese e-mail culture tends to be more flowery than Anglo (“Espero que esteja bem”, longer greetings, indirect requests). North-American style is direct, value-first, and skim-friendly. For B2B, prefer short, value-led copy with a P.S. line driving the CTA — the P.S. is the second most-read line after the subject. For B2C, lead with emotion and benefit, and personalise heavily.

Subject lines decide everything

Open rate hinges on the subject line. Keep it under 6 words, personalise (first name, company name), avoid spammy words (FREE, URGENT, all caps) and skip emojis in formal contexts. A/B-test variants in Mailchimp, HubSpot, Apollo.io or Lemlist. For cold email, mention a mutual connection (“Maria sugeriu”) to lift opens by 30-50%.

Compliance and tooling

• LGPD (Brazil), GDPR (EU), CAN-SPAM (US), CASL (Canada) — mandatory unsubscribe link, legal basis for processing, identification of sender.
• Sales cadences: Apollo.io, Outreach.io, Lemlist, HubSpot Sales, Salesloft.
• Deliverability: SPF, DKIM, DMARC set up on the sending domain.
• Warm-up: Lemwarm, Mailwarm — new domains need 2-3 weeks of ramp-up.

Anti-patterns

Avoid “I hope this email finds you well” — it screams template. Avoid 300-word paragraphs — people skim. Avoid generic CTAs (“Let me know what you think”) — ask for a specific 15-minute slot or a yes/no decision. Avoid copy-pasting the LLM output verbatim — replace 2-3 sentences with your own voice.

FAQ

Can the LLM produce a professional-grade email? Yes — with RTCF and a clear framework (AIDA, PAS) it consistently produces shippable drafts. Always customise 2-3 sentences before sending.

Should I let the model translate PT ↔ EN? Translation is solid for content, but always reread the tone — Anglo directness can sound rude in PT, and Brazilian politeness can sound subservient in EN.

Is it OK to send the output verbatim? Better not. Recipients now spot “ChatGPT prose” (em-dashes, balanced two-clause sentences, predictable transitions). Always rewrite the opening line.

Which model writes the best emails? Claude tends to handle tone and nuance better; GPT-4o is faster and slightly more concise. For cold outreach, both work — the prompt and your offer matter much more than the model.