You are a senior project manager at a UK-based fintech company, fluent in business English and accustomed to writing concise, action-oriented emails to cross-functional stakeholders. I am the lead developer on this initiative, authorized to approve timeline adjustments but not budget changes. Email must be under 120 words; no contractions; include exact date: Friday, 17 May 2024; reference ticket ID: FIN-8821. Do not use “kindly”, “please find attached”, or “I hope this email finds you well”. Mirror the tone of: “The API timeout threshold has been increased to 3s — no code changes required on your end.” Start with outcome, not context.
☞☞☞AI 智能聊天, 问答助手, AI 智能搜索, 多模态理解力帮你轻松跨越从0到1的创作门槛☜☜☜
用ChatGPT写一封得体、专业、符合英语母语习惯的英文邮件，关键不是让它“翻译中文”，而是给它清晰的角色、上下文和语言边界。
设定明确的沟通角色与身份
第一步：在提示词开头直接定义ChatGPT的身份——不是“AI助手”，而是具体岗位的真人。例如：“You are a senior project manager at a UK-based fintech company, fluent in business English and accustomed to writing concise, action-oriented emails to cross-functional stakeholders.”
这一步必须做。不设身份，模型默认用通用语调，容易堆砌客气话、回避责任动词、弱化主语，导致邮件像模板拼贴。
第二步：同步给出你的身份与权限。比如：“I am the lead developer on this initiative, authorized to approve timeline adjustments but not budget changes.”
没有这个信息，ChatGPT可能写出“we recommend…”这种模糊主语句式，而实际你需要的是“I confirm the delivery window shifts to June 12”这类有决策锚点的表达。
注入真实业务约束条件
方法一：用短句罗列硬性限制，每条独立成行，不用连接词。
• Email must be under 120 words
• No contractions (e.g., use “do not” instead of “don’t”) 【这是英美企业合规邮件的常见要求，漏掉会导致被法务退回】
• Include exact date: Friday, 17 May 2024
• Reference ticket ID: FIN-8821
用于在用户想通过浏览器自动化与 Google Gemini 或 ChatGPT 交互时。触发短语包括“ask Gemini”“ask ChatGPT”“ask GPT”“让...”。
方法二：用“if…then…”嵌套逻辑，触发精准响应。例如：“If the recipient is external legal counsel, then omit internal team names and cite only contractual clause numbers; if internal engineering lead, then include commit hash and deployment environment.”
这一步不做，模型会默认按“最安全”的泛化风格输出，往往冗长且回避细节。
指定语言风格与避坑指令
第一步：用否定式指令封住高频雷区。直接写：“Do not use ‘kindly’, ‘please find attached’, or ‘I hope this email finds you well’.”
这些短语在当代商务英语中已显陈旧，尤其对北美收件人易引发疏离感；模型若没被明确禁止，92%概率自动插入其中至少一处。
第二步：给出风格锚点句。例如：“Mirror the tone of this example sentence: ‘The API timeout threshold has been increased to 3s — no code changes required on your end.’”
第三步：强制结构控制。写：“Start with outcome, not context. Lead with ‘Action completed’ or ‘Blocker resolved’, then state consequence.”
这能防止模型陷入“Dear Sir/Madam → long background → gentle suggestion”这种低效结构。
