According to a Harvard study, knowledge workers spend an average of 28% of their work time writing and answering emails. Add reports, presentations, and internal communication on top. This is precisely where AI delivers the fastest return on investment — often on day one.
Emails are the biggest quick-win area. Three scenarios where AI helps immediately:
1. Generate replies Copy the incoming email and prompt:
"Reply to this email professionally and warmly. Confirm the Thursday appointment, request the agenda in advance, and suggest a 15-minute pre-call."
2. Summarize long threads
"Summarize this email thread: Who said what, what decisions were made, what action items are open?"
3. Draft difficult messages
"Write a diplomatic decline for a meeting request. Reason: schedule conflict. Suggest two alternative times next week."
📖 Definition: Prompt engineering for emails means clearly providing the AI with context (incoming email), the desired tone (professional, diplomatic, direct), and key messages (confirmation, decline, question).
💡 Tip: Save your best email prompts as templates. In Claude, use Projects for this; in GPT-5, use Custom Instructions. After a week, you'll have a personal prompt library.
For reports and summaries, Claude Opus 4.6 is particularly well-suited because its long context accommodates extensive source data.
| Document Type | Prompt Template | Best Tool |
|---|---|---|
| 📊 Management Summary | "Create an executive summary from this data (max 300 words). Audience: C-level." | Claude Opus 4.6 |
| 📝 Meeting Minutes | "Structure these notes: Participants, topics, decisions, action items with deadlines." | GPT-5, Claude |
| 📈 Quarterly Report | "Analyze these KPIs and write a quarterly report with trend analysis." | Claude Opus 4.6 |
| 📋 Project Status Report | "Create a status report: progress, risks, next milestones." | GPT-5, Claude |
🏢 Real-world example: A project manager uses Claude Opus 4.6 to create weekly status reports from Jira exports and Slack summaries. Time saved: 2.5 hours per week. Quality improved because the AI structures consistently and never forgets details.
Modern voice-to-text tools have revolutionized transcription:
A typical workflow: Record meeting → Otter.ai transcribes → Claude summarizes the transcript and extracts action items → Result posted to Slack.
⚠️ Caution: Inform all participants before recording a meeting. In many jurisdictions, recording without consent is illegal. Check the policies with your legal department.
AI translations have reached professional quality. GPT-5, Claude Opus 4.6, and Gemini 3.1 don't just translate literally — they adapt tone, technical terms, and cultural conventions:
"Translate this business report into English. Maintain the formal tone. Adapt date formats, currencies, and cultural references for a US audience."
For recurring translations, a translation memory pays off: Collect verified translations of technical terms in a document and provide it to the AI as reference.
AI texts aren't perfect. Always check these five points:
🔑 Remember: AI-generated texts are a draft, not a final product. The last mile — fine-tuning, fact-checking, and tonality — remains a human task. But that last mile takes 10 minutes instead of 2 hours.
🎯 Exercise: Take three emails from your inbox that need a reply. Generate the replies with Claude Opus 4.6 or GPT-5. Measure the time and compare it with your usual processing time.
Next lesson: AI for Data Analysis and Visualization — Your Personal Data Analyst