Professional writing is a high-leverage skill: one clear email can unblock a project, one strong letter can win a partnership, and one well-structured resume can land interviews. Generative AI tools (ChatGPT, Claude, Gemini, etc.) can help—but the best results come from using AI as a drafting and revision partner, not a “write it for me” button.
This guide shares a practical workflow, reusable prompts, and concrete examples for writing better emails, letters, and resumes—while keeping your voice and staying on the right side of privacy and honesty.
The mindset: AI as a writing system, not a magic trick
AI is great at:
- Generating drafts from bullet points
- Adapting tone (formal, friendly, concise)
- Improving clarity and structure
- Spotting gaps (missing context, unclear asks)
- Tailoring content to an audience
AI is risky at:
- Inventing details (dates, achievements, metrics)
- Overconfident claims (“expert in…”) you can’t defend
- Producing generic phrases that hiring managers spot instantly
- Leaking sensitive information if you paste confidential text
A reliable approach is: you provide the facts and constraints; AI provides options and polish; you verify and finalize.
A simple 4-step workflow that works across emails, letters, and resumes
1) Start with “truth bullets”
Before you prompt, write 5–10 bullets containing only facts you’re comfortable sharing:
- Who is the audience?
- What’s the goal?
- What are the key details (dates, numbers, scope)?
- What constraints exist (tone, length, policy, legal)?
- What action do you want next?
This reduces hallucinations and keeps the draft aligned with reality.
2) Ask for multiple variants
One draft is rarely perfect. Ask for 2–4 options with different tones or structures. You’ll usually combine the best parts.
3) Run a “risk check” pass
Have the model list any claims that sound unverifiable or too strong, and anything that might be inappropriate for the audience.
4) Final human edit
Read out loud. Remove filler. Add a human detail (a specific project, a shared context, a natural sign-off). If it sounds like a template, it is.
Prompt patterns you can reuse
These prompt templates work in most LLMs. Replace bracketed text with your details.
Template A: Draft from bullets
“Write a [email/cover letter/resume bullet] based on these facts. Keep it [tone], [length]. Do not invent details. If something is missing, ask me questions.
Facts:
- [bullet]
- [bullet]
Audience: [who] Goal: [what outcome] Call to action: [what you want next]”
Template B: Rewrite for clarity + tone
“Rewrite the text below to be clearer and more concise. Keep my meaning and do not add new facts. Produce two versions: (1) warm and collaborative, (2) direct and executive.
Text: [copy/paste]”
Template C: QA + red flags
“Review this draft for:
- unclear asks,
- missing context,
- overly strong claims,
- possible misinterpretations. Then propose minimal edits.”
Template D: Audience adaptation
“Adapt this message for [audience], considering they care about [priorities]. Keep it under [word count].”
Use case 1: Better emails (faster, clearer, more effective)
Most professional email problems aren’t grammar—they’re unclear goals. AI helps most when you give it the goal and constraints.
Example: Project update email
Your truth bullets
- Audience: product + engineering stakeholders
- Goal: align on scope change and next steps
- Facts: API integration is blocked by vendor rate limits; mitigation is caching + batching; adds ~5 engineering days; still on track for end-of-month if we start today
- Ask: approve scope change and confirm who owns vendor escalation
Prompt (Template A) “Write a project update email based on these facts. Tone: calm, accountable, no blame. Length: 150–200 words. Do not invent details. End with clear next steps and owners.
Facts:
- API integration is blocked by vendor rate limits.
- Proposed mitigation: caching + batching.
- Effort: ~5 engineering days.
- Timeline: still on track for end-of-month if we start today.
- Need: approve scope change; confirm owner for vendor escalation.
Audience: product + engineering stakeholders Goal: align on scope change and next steps Call to action: approve mitigation plan + assign escalation owner”
Pro tip: Ask for a “one-screen” version
Add: “Make it skimmable: short paragraphs, 3 bullets max.” This yields messages people actually read.
Pro tip: Create a personal style guide
Ask AI to infer your style from 2–3 emails you wrote (after removing confidential info), then save a short style guide: preferred greeting, sign-offs, sentence length, and whether you use bullets.
Use case 2: Letters (cover letters, recommendations, outreach)
AI can help you avoid generic language, but only if you feed it specifics—especially for cover letters.
Cover letter approach that doesn’t sound templated
Instead of asking for a full letter immediately, generate the core story first:
Prompt “Based on my bullets, create:
- a 2-sentence ‘why this role’ pitch,
- three achievement stories in STAR format (Situation/Task/Action/Result) using only provided facts,
- a list of 5 specific company-facing questions I could ask in an interview. Do not invent metrics.
Bullets:
- [your facts]”
Then ask the model to assemble a letter using the strongest story.
Recommendation letters: be explicit about constraints
If you’re writing a recommendation, add constraints like: relationship length, what you directly observed, and what you cannot claim.
Prompt snippet “Write a recommendation letter. Only include what I observed firsthand. Avoid exaggeration words like ‘best ever’ unless supported.”
Use case 3: Resumes that pass the “so what?” test
Resumes are where AI can help most with structure and impact phrasing—but it can also produce the most fluff. The key is to anchor every bullet in measurable outcomes or concrete scope.
Start by converting responsibilities into outcomes
Try this transformation prompt:
“Here are my raw job notes. Convert them into 6 resume bullets using the format: action verb + what + how + result. If a result metric is missing, suggest placeholder options like ‘reduced by X%’ but label them as placeholders for me to fill.
Raw notes:
- [notes]”
Example: Turning vague into strong
- Vague: “Worked on dashboards for leadership.”
- Stronger: “Built executive dashboards in [tool] to track [KPIs], enabling weekly decision-making across [teams].”
Notice the stronger version still requires truth: tool, KPIs, and scope must be accurate.
Tailor to job descriptions—without keyword stuffing
AI is useful for mapping your experience to a posting:
Prompt “Compare my resume bullets to this job description. Identify:
- overlapping skills to emphasize,
- missing but relevant experiences I might have (ask me questions),
- keywords to incorporate naturally. Then propose edits to 5 bullets max.
Resume bullets: [bullets]
Job description: [text]”
This keeps changes focused and reduces the “AI wrote my whole resume” vibe.
A practical “anti-bot” checklist
Before sending or submitting, scan for common AI tells:
- Overuse of “I am excited to…” (replace with a specific reason)
- Generic claims (“results-driven,” “detail-oriented”) without evidence
- Long, balanced sentences with no punch
- Too-perfect politeness in emails that should be direct
A simple fix: add one specific detail only you would know (a project constraint, a shared meeting, a real metric, a named tool).
Privacy and ethics: what not to paste into a chatbot
Be careful with:
- Client names, proprietary project details, internal metrics
- Unreleased financials, HR issues, medical info
- Anything covered by NDA
Options:
- Redact: replace names with placeholders (e.g., “Client A,” “Product X”).
- Summarize: provide high-level bullets instead of raw documents.
- Use enterprise tools: if your company provides an approved AI environment with proper data handling.
Ethics for resumes and letters is simple: AI can help you write; it cannot help you lie. If you can’t defend a claim in an interview, remove it.
A mini-tutorial: build your own “email copilot” prompt
Save this as a reusable prompt in your tool of choice:
“Act as my writing assistant for professional communication.
My preferences:
- Tone: [friendly/direct], minimal fluff
- Structure: short paragraphs + bullets when helpful
- Always include a clear ask and deadline when relevant
- Never invent facts; ask questions if needed
Task:
- Draft an email based on my bullets.
- Provide a subject line.
- Provide a 1-sentence follow-up message if no response in 48 hours.
Bullets: [insert]”
This turns AI into a consistent system rather than a one-off tool.
What’s next: bring a real draft to the NH AI Meetup
If you want to level up quickly, bring:
- One real email you send often (status updates, meeting requests, customer replies)
- One resume bullet that feels weak
- A job description you’re targeting
Then practice iterative prompting: draft → critique → revise → finalize. The skill isn’t “prompt engineering” so much as clear thinking, structured facts, and deliberate editing.
Used well, AI won’t replace your voice—it will amplify it, helping you communicate with more clarity, confidence, and speed.