What you'll get from this lecture

What you'll need

Open any AI tool — Claude, ChatGPT, Gemini, Copilot. All PROMPT blocks below can be pasted and run.

Important: For business information, always follow your company's AI-usage guidelines. Customer / patient / unpublished data should never be put into AI not approved by the guidelines. This lecture assumes non-confidential practice data.

FLOW 1 ── Drafting email replies in 5 minutes

What used to take 10 minutes per email becomes 3 minutes when written with AI.

Step 1.1Share the received email with AI

Select one email from your inbox and paste it into the AI.

▶ Your action
We'll draft a business email reply together. Here is the received email: --- (paste the email) --- Summarize the gist of this email in 3 lines or fewer: 1. What is the other party asking? 2. Deadline / conditions 3. Tone required for the reply (formal / casual / urgent, etc.)
Expected: AI returns the gist in 3 lines. You can check whether your understanding matches.
Step 1.2Hand the AI the core of your reply, as bullets

Before having AI write, hand it what you want to say, as bullets. This is the biggest trick for "controlling" AI.

▶ Your action
My intended message is the following 3 points: - (example) Basically agree with the proposal - However, the budget condition requires discussion - Would like to discuss details at next Thursday's meeting Based on this, write a reply email of about 200 characters matching the recipient's tone. Also suggest a subject line.
Expected: A polished draft. ~80% should be ready to use.
Step 1.3Final adjustment (the human part)

Read the AI output and check whether it sounds like you. Fix 2–3 words that feel off or expressions you wouldn't use. Done.

Time: 1–2 minutes for the email body + 2–3 minutes with AI = 3–5 minutes total.

FLOW 2 ── Finishing meeting minutes during the meeting

The "I'll compile minutes later" task eats your time. Finish them during the meeting with AI.

Step 2.1Transcribe the meeting audio to text

During the meeting, open a notes app on your PC (Notion, Apple Notes, OneNote). Many tools have voice input that transcribes in real time. It doesn't have to be perfect; AI can fix typos afterward.

For in-person: record, then transcribe later with Whisper or ChatGPT's voice feature.

Step 2.2Have AI shape it into "minutes"

Hand it to AI immediately after the meeting ends.

▶ Your action
The following is a meeting transcript: --- (paste transcript) --- Convert this into minutes in the following format: # Meeting Minutes - Date/Time: - Attendees: - Agenda: ## Decisions - (bullet list; each item with owner and due date) ## Under discussion - (bullet list; make the open questions explicit) ## Action items before next meeting - (Owner / item / due date) ## Open notes (impressions, unstructured) - ...
Expected: Clean minutes in 30 seconds. Manually correct only AI-error-prone things like attendee names.
Step 2.3Distribute within 5 minutes of meeting end

Review and share with attendees. Within 5 minutes of meeting end — this alone raises your workplace trust significantly.

What took 30–60 minutes becomes 5–10 minutes.

FLOW 3 ── Summarizing long documents in 5 minutes

Papers, internal documents, regulations — grasp 20–30 page documents as 5-line essentials.

Step 3.1Define the "purpose" of the summary first

Output of a summary depends on what you want to know. Asking vaguely "summarize this" returns a vague summary.

▶ Your action
Summarize the following document. 【My purpose】 (example) I want to judge whether this paper can be used in our company's promotional materials for Drug A. Extract: - Conclusion - Study design - Efficacy data - Safety data - Material for judging citation suitability 【Document】 --- (paste document) ---
Expected: A structured summary aligned with your purpose. The moment you read it, your next action is clear.
Step 3.2Ask the AI about blind spots

Summary alone is insufficient. Ask the AI "what am I not understanding?"

▶ Your action
For this summary, list 3 questions I should check next. Also point out areas likely to be my blind spots.
Expected: Questions like "What were the subgroup analysis results?", "What were the control group selection criteria?" — perspectives you wouldn't have asked yourself. This is the knack for using AI as a "dialogue partner".

Combining the three ── "morning 30-minute routine"

Combine the three FLOWs into a 30-minute routine to run every morning.

TimeTaskFLOW
0–5 minProcess morning unread emails with AIFLOW 1
5–15 minAI-shape and distribute yesterday's meeting minutesFLOW 2
15–25 minAI-summarize papers/docs to read today → deep-read only relevant onesFLOW 3
25–30 minList today's priority tasks with AIApplied

What used to take 1.5–2 hours fits in 30 minutes. Over 22 working days per month, you recover 22–33 hours. That's "half a day to a full day" of work buffer.

Failure cases ── what not to do

Save your templates

Save the PROMPT blocks from this lecture as your "prompt library". Notion, Apple Notes, plain text — any. Next time, just paste. This is the fastest path to mastery.

In closing

AI is not a magic wand. But it is a tool that reliably returns 30 minutes a day to you.

The three flows here are simple. Because they're simple, you can start tomorrow. Sustain for a week. If you keep it for a month, you will not want to give it up.

The shift from "I'm not good with AI" to "I work with AI every day" happens through this kind of small accumulation. Hope this lecture is your first step.