PromptingBasics

Email Prompts

By
Dan Lee
Dan Lee
Dec 20, 2025

Writing Email with AI (Without Sounding Like a Robot)

AI can write emails ridiculously fast—but only if you give it the right constraints. Otherwise you’ll get the classics: overly formal greetings, generic fluff, and paragraphs that say a lot without saying anything.

This guide gives you a simple structure, plus a few reusable prompt templates you can copy/paste for real work (sales, recruiting, exec assistants, engineering, customer support, leadership).

The Email Formula

For 90% of emails, you want: Context → Ask/Action → Details → Close. If the model doesn’t know those, it will ramble.

The “CIDO” Email Prompt

A reliable email prompt includes four pieces:

  • Context: who you are, who they are, what happened
  • Instruction: what you want the email to do (schedule, confirm, apologize, follow up)
  • Data: the facts (dates, links, pricing, constraints, prior messages)
  • Output indicator: subject line, length, tone, structure

Example 1: Executive Assistant (Reschedule + Options)

Text
Context: I’m an executive assistant writing on behalf of our CEO. We need to reschedule a 30-minute meeting with an investor due to a conflict.
Instruction: Draft a concise rescheduling email that maintains trust and makes it easy to pick a new time.
Input Data:
- Original meeting: Tue 2:00–2:30pm ET
- Conflict: last-minute board prep
- New options: Wed 10:00–10:30am ET, Thu 3:00–3:30pm ET, Fri 11:30am–12:00pm ET
Output Indicator:
- Provide 2 subject lines
- Email body <= 110 words
- Tone: warm, professional, respectful
- End with a single clear question asking them to choose a slot

Why it works: it forces brevity, provides options, and creates a low-friction next step.

Example 2: Engineer → PM (Status Update That’s Not Chaos)

Text
Context: I’m a software engineer sending a project status update to a PM and stakeholders.
Instruction: Write an email that summarizes progress, flags risks, and proposes next steps.
Input Data:
- Shipped: auth refactor (done), logging improvements (done)
- Blocker: vendor API rate limits causing flaky tests
- Decision needed: approve $250/mo for higher rate limit tier OR accept reduced test frequency
- Deadline: feature demo next Friday
Output Indicator:
- Subject line + email body
- Body structure: 3 bullets for Progress, 2 bullets for Risks, 2 bullets for Next Steps
- Tone: crisp, no jargon, no blame
- <= 160 words

This produces an email stakeholders can skim and respond to quickly.

Ask for Two Versions

If the email feels high-stakes, request “Version A (direct)” and “Version B (more diplomatic).” Then you pick the right one for the room.

Copy/Paste Templates (Quick Wins)

1) Follow-up after no reply

Text
Draft a polite follow-up email. Include (1) a one-line reminder, (2) a simple CTA question, and (3) a friendly close.
Keep it under 90 words. Tone: calm and professional.
Context: [who/what]

2) Customer support resolution

Text
Write a customer support reply that (1) confirms the issue, (2) apologizes briefly, (3) states the fix, and (4) gives the customer one next step.
Tone: helpful, confident. Avoid excessive apology. <= 140 words.
Issue details: [paste]

3) Recruiting outreach

Text
Write a recruiting outreach email to a candidate for [role]. Personalize using the profile notes.
Include: why them (1 sentence), role impact (2 bullets), and a 15-min chat CTA.
Tone: human, not spammy. <= 130 words.
Profile notes: [paste]

Takeaway

Great AI email writing isn’t about fancy prompts—it’s about clear intent + constraints. Give the model context, the exact job to do, the facts to use, and the output shape you want (subject + tone + length + structure). Do that, and AI stops sounding like a template machine—and starts sounding like you, but faster.

Dan Lee

Dan Lee

DataInterview Founder (Ex-Google)

Dan Lee is an AI tech lead with 10+ years of industry experience across data engineering, machine learning, and applied AI. He founded DataInterview and previously worked as an engineer at Google.