Table of Contents
Artificial Intelligence Translates the Language of Animals
Kim Kyung-jin, Attorney at Law
The Story of AI Learning to Listen to Whales, Dolphins, Birds, and Bees
Twelve chapters on how AI listens to dolphins, sperm whales, humpback whales, birds, and bees to find rules in their sounds, and what this technology means for its risks and for animal rights. Written in simple sentences a child can read, with verified sources in every section.
Table of Contents
AI Deciphers Ancient Scripts
Kim Kyung-jin, Attorney at Law
Ancient Records Revived by AI
In twelve chapters, this book explains how AI revives records once unreadable, from burned scrolls and wooden slips buried in mud to broken clay tablets. It covers virtual unrolling at Herculaneum, virtual collation of oracle-bone texts, reading Silla wooden tablets, computational analysis of undeciphered scripts, and multispectral archives, with verified references for each chapter.
Table of Contents
Artificial Intelligence for New Materials Design and Rocket Propulsion Engineering
Kim Kyung-jin, Attorney at Law
AI Potentials, Self-Driving Laboratories, and Physics-Informed Machine Learning (PIML)
Ten chapters on how artificial intelligence is changing new materials and rocket propulsion: atomic simulation, generative models, self-driving labs, high-temperature alloys, metal 3D printing, combustion, cooling design, and engine diagnosis and control. Written without equations, with verified sources in every chapter.
AI Library
The Age of Autonomous Scientific Discovery
Kim Kyung-jin, Attorney at Law
AI Scientists and Self-Driving Labs
This book follows how AI scientists and self-driving labs are changing the way science generates and verifies claims. It covers literature-based discovery, natural-language protocols translated into robot commands, multi-agent research systems, closed-loop laboratories, materials search, the verification gap, chains of evidence, research harnesses, journal ethics, and legal responsibility.
AI Library
A New Era of Life Sciences Opened by Artificial Intelligence
Structural Proteomics, Genomic Foundation Models, Autonomous Laboratories, and Global Governance
Kim Kyung-jin, Attorney at Law
This book is a research volume compiled with artificial intelligence. A human selected the materials and structured the work, while AI models drafted the sentences and cross-checked the facts.
AI Library
The Double Structure of Digital Sovereignty
Europe’s Departure from Palantir and the Chains of American Big Tech
Kim Kyung-jin, Attorney at Law
This is a record of 2026, when European intelligence agencies and defense ministries began removing analytics tools from America’s Palantir. It covers the replacement decisions made by France’s General Directorate for Internal Security (DGSI), Germany’s Federal Office for the Protection of the Constitution (BfV), and the Netherlands Ministry of Defense; the incident in which US export controls severed an ally’s ac…
New English Edition
Artificial Intelligence in Horticulture
Kim Kyung-jin, Attorney at Law
Across five chapters and ten sections, this book examines computer vision for crop diagnosis, harvesting robots and autonomous field systems, smart greenhouses and digital twins, precision irrigation and supply-chain quality control, high-throughput phenotyping, and predictive breeding.
New English Edition
Artificial Intelligence in Food Crop Agriculture
Kim Kyung-jin, Attorney at Law
Across six chapters and eighteen sections, the book examines digital agricultural infrastructure, remote sensing, crop diagnosis, yield forecasting, precision irrigation, genomics, molecular breeding, agricultural robotics, climate-smart agriculture, and global food security.
New English Edition
The Future of Forestry and Agroforestry
Kim Kyung-jin, Attorney at Law
Driven by Artificial Intelligence and Digital Innovation
Across five chapters and fifteen sections, the book follows satellites, drones, LiDAR, digital twins, forest-specific language models, wildfire and pest forecasting, forestry robotics, agroforestry, timber traceability, and forest carbon markets.
New English Edition
Smart Livestock Farming: AI Enters the Barn
Kim Kyung-jin, Attorney at Law
Sensors listen, cameras watch, and artificial intelligence helps farmers decide.
Across five chapters and fifteen sections, the book follows precision livestock farming from animal health and reproduction to robotic milking, virtual fencing, digital twins, methane reduction, welfare, and data ownership.
Table of Contents
Han Dong-hoon, Busan Buk-gu Gap: A Record of the 100 Days Before and After the Election (Mar. 26-Jul. 3, 2026)
Kim Kyung-jin
Table of Contents and 13 sections
From March 26 to July 3, 2026, this record follows the spring after expulsion, the Busan Buk-gu Gap by-election, victory as an independent, and the first bill submitted in the National Assembly.

Table of Contents
Artificial Intelligence and Medicine
Kim Kyung-jin, Attorney at Law
AI in clinical care, hospitals, education, and research
AI in medical imaging, risk prediction, treatment planning, hospital operations, education, and research, with patient safety, privacy, and accountability.
[AI Library] Chapter 5: Daily Operations and Communication Management
Claude Cowork and Agent Utilization Manual
Chapter 5: Daily Operations and Communication Management
Kim Kyung-jin
After a meeting ends, memories diverge about who was supposed to do what. Each morning, opening your inbox reveals an important message buried among advertisements. Five minutes before a meeting, you confirm the other person's name. The night before a business trip, you pack your suitcase and try to remember where your passport is. These tasks are not the work itself. They are the work needed to do the work,what we call meta-tasks.
The four tasks covered in this chapter represent the area where Cowork shows the most immediate impact.
They are: analyzing meeting transcripts, organizing email and drafting replies, preparing meetings by syncing with calendar, and planning business trips.
They share a common trait: all are repetitive, structured, and require speed more than precision. When done by a person, they take thirty minutes; delegated to Cowork, three minutes. That difference reshapes your day.
Before we begin this chapter, one point deserves attention. As of March 2026, Gmail, Google Calendar, and Google Drive connectors (the connections that link Cowork to external services) are available in Claude's desktop app chat tab, and can also be connected from the Cowork tab.
However, since Cowork is still in research preview (a trial phase before official release), the scope of connector support and how they work may change with updates. The exercises in this chapter are based on official documentation as of March 2026.
1 Recording Meetings and Analyzing Transcripts
People in the same meeting hear the same words, yet their memories of who was supposed to do what diverge. In a thirty-minute meeting, only two or three things actually get decided, while the remaining twenty-five minutes of conversation blur that decision. In the past, the junior staffer would type throughout the meeting, then afterward piece together notes from memory. Context fell away. Critical commitments were missed. Responsibility became unclear.
Tools like Fireflies, Tana, and Clova Note have solved half this problem: converting speech to text. But a one-hour meeting's transcript fills fifteen to twenty pages of A4 paper.
Reading through it all and extracting the core content becomes another task in itself. Cowork handles precisely this second half: taking raw text and building structure from it.
A. Basic Use: Summarizing a Transcript and Extracting Action Items
Start with the simplest form: hand a meeting transcript text file to Cowork and receive a summary and follow-up list.
[How to do it]
① Open Claude's desktop app and click the "Cowork" tab in the left sidebar.
② Designate a working folder. Create a folder named "Meeting Minutes" on your desktop beforehand, place the transcript text file (.txt) in it, then select this folder from the top of the Cowork screen.
③ Enter the following prompt.
"Please read the meeting transcript file in this folder and organize these three things:
First, summarize the entire meeting in ten lines or fewer. Avoid vague phrasing like 'it was discussed'; instead write decision-focused language like 'Option A was adopted and Option B was deferred.' Second, organize the decisions made in the meeting as a numbered list. Third, create a table for follow-up actions (action items). Include columns for owner, task, deadline, and priority. Save the results as a Word document in this folder. Name it 'Meeting Summary_Date.'"
④ Claude displays a plan. Steps appear on screen: "I will read the transcript, create a core summary, separate the decisions, build the action item table, generate the Word document." Review the plan and approve execution.
⑤ Claude executes the work. Progress displays stage by stage. When complete, a Word document appears in the "Meeting Minutes" folder.
The key here is separating summary from action items. What was agreed in the meeting and who must do what by when to carry out that agreement are separate pieces of information. Claude distinguishes between the two.
[To know] What format should the transcript file be?
Cowork can read .txt, .md, .docx, and .pdf files. When exporting transcripts from Fireflies or Clova Note, save as a text file (.txt) or Word file (.docx). If you include the meeting date and subject in the filename, Claude can automatically match the document title, which is convenient.
B. Applied Example: Changing Output Format and Creating an Excel Tracking Table
If you received one Word document in the basic version, this time you divide outputs by use case: creating both a summary document for team sharing and an Excel tracking table for project management simultaneously.
A Word document is formatted for readers. Executives, managers, and clients look at the summary structure first. Excel, by contrast, excels at tracking. In organizations with many action items, you need a table separate from the notes, one you can filter by owner and sort by deadline.
[How to do it]
① Reuse the "Meeting Minutes" folder you created before. Modify the prompt as follows.
"Analyze the transcript in this folder and create two files. First, a Word document for executive reporting. Structure it as: Meeting Overview, Attendees, Key Decisions, Risk Factors, Next Meeting Agenda. Keep it to one A4 page maximum. Second, an Excel file for tracking follow-ups. Include these columns: Meeting Date, Topic, Owner, Task, Deadline, Priority, Status, Notes. Highlight items marked 'Urgent' in the Priority column with red cell backgrounds. Save both files in this folder."
② Claude displays a plan. This time, the plan shows generating two files in parallel. For complex work, Claude may spawn sub-agents (auxiliary AIs handling subsidiary tasks) to work simultaneously.
③ After completion, check the folder. Both the Word document and Excel file are saved together.
An Excel tracking table amplifies the impact of recurring weekly meetings. If you accumulate action items in the same format each week, after a month you can see which topics have cycled for weeks without resolution, which owner's items keep rolling forward. The notes become more than a record; they become a tool that measures organizational execution.
C. Real-World Application: Saving Recurring Meetings as Skills and Automating Them
If you have regular meetings,like a weekly Monday check-in or a first-week-of-month executive meeting,you need not enter the same prompt each time. Save a working workflow as a skill (reusable work instructions stored in Cowork).
[How to do it] ① If the transcript analysis work you just ran satisfied you, enter this in the same Cowork session: "Please save the meeting transcript analysis workflow I just ran as a skill. Name the skill 'Weekly Meeting Post-Processing.'"
② Claude organizes the work instructions into a Markdown file and saves it in the skills folder.
③ When next week's transcript is ready, place it in the folder and enter just one line in Cowork: "Please apply the 'Weekly Meeting Post-Processing' skill to this week's transcript."
Go one step further and create it as a scheduled task (a recurring job that runs at a set time). Type /schedule in the Cowork input field and a scheduling screen appears. Set it to "Every Monday at 9 a.m., check the 'Meeting Minutes' folder. If there is a new transcript file, run the 'Weekly Meeting Post-Processing' skill." When you arrive Monday mornings, the organized meeting document is already in the folder.
However, for a scheduled task to run, your computer must be on and Claude's desktop app must be open. It does not work when your laptop is closed.
Organizations with frequent customer meetings can use this to accumulate meeting history by customer. Gather meeting transcripts in a subfolder labeled "Customer A." Before the next meeting, request "Extract only open items from all meeting notes in this folder" to see past discussion points and remaining tasks at a glance.
One point must be stated plainly. Claude excels at converting spoken words to text and identifying surface-level causality.
But Claude can misread subtle atmospheres in a conference room, implicit disagreement conveyed through silence, or joking sarcasm as earnest proposals. Whether Claude's action item table fully captures the truth of the conversation is the human's responsibility. The person freed from the labor of note-taking now plays the editor's role, filling in the context the machine missed.
d If You Run Into These Issues
(1) I uploaded a transcript file, but Claude says "File not found." Check that the correct folder is selected at the top of the Cowork screen. If you don't designate a folder, Claude cannot access the files. Even if you have subfolders within the folder and placed the file inside one, access may fail. Try placing the file at the top level of the designated folder.
(2) The summary is too long, or it looks like the original text was copied almost verbatim. Specify length limits in the prompt. When you include specific numbers like "within 10 lines" or "one A4 page or less," Claude adjusts the compression level. The instruction "focus on decisions" is also important. Without this one line, Claude will try to describe the discussion process as well.
(3) The owner or deadline for action items is inaccurate. When the transcript mentions the owner's name unclearly (for example, "if your side could handle it..." or "someone please take a look"), Claude may guess and assign it incorrectly. You must review the output, fix any errors, and then share it with your team. A document Claude creates is a draft, not a final version.
2 Email Inbox Organization and Reply Automation
When you open your laptop in the morning, dozens of emails have accumulated overnight. Half are newsletters, a few have your name only on the CC line, and only a few need immediate action. The problem is the time it takes to find those few messages. As you open emails one by one, judge their importance, and close them, your high-focus morning hours slip away.
When you link Claude with the Gmail connector, you can structure this process. Claude retrieves a list of unread emails, then analyzes the sender, subject, and first part of the body to categorize them by importance. For important emails, it drafts replies that match your tone of voice.
Understanding exactly what the Gmail connector does eases your mind. According to Anthropic's official documentation, the Gmail connector can search for, read, and draft emails, but it does not send mail directly on your behalf. Drafts are saved in your Gmail account's draft folder, and you must send them yourself. This safeguard prevents Claude from accidentally sending the wrong reply.
a Basic Use: Categorizing Unread Emails by Importance
We start with the process of connecting the Gmail connector for the first time.
[Follow Along] 1. Open Settings in the Claude desktop app.
2. Find the Connectors section and click "Browse connectors." Locate Gmail in the list and click "Connect."
3. The Google sign-in window appears. Sign in with the Gmail account you want to use and approve the access permissions.
4. Once the connection is complete, return to the Cowork tab. Enter the following prompt.
"Check all unread emails in my Gmail and categorize them into these three levels.
Urgent: direct customer inquiries, contracts, payments, mail from courts or government offices. Standard: internal work requests, schedule coordination, colleague questions. Reference: newsletters, advertisements, CC'd emails, automated notifications. Organize the classification results in a table. Include the sender, subject, one-line summary, and importance level in the table."
5. Claude accesses Gmail and scans the emails. A popup may appear asking for approval to access each email. Click "Allow."
6. Claude displays the classification results in a table on screen. A briefing is complete, such as "Three emails to check immediately today, five emails to handle today, and twelve emails to just review."
You can change the classification criteria to fit your work. For a lawyer, court mail and direct customer inquiries come first; for a shop owner, customer complaints and shipping-related emails come first. The more specific your work context in the prompt, the higher the classification accuracy.
b Application: Drafting Replies in Your Brand Tone
Once categorization is done, it's time for replies. The biggest concern with reply automation is that it sounds like a machine wrote it. Cowork offers a solution: teaching Claude your past emails to create a profile of your tone and writing habits.
[Follow Along]
1. First, teach Claude your writing tone. In Cowork, enter the following.
"Read 20 emails I wrote directly over the past 30 days in my Gmail sent folder, and create my email tone profile. Analyze greeting patterns, sentence length, frequently used phrases, signature style, and formality level. Save the result as a file named my_email_tone.md in this folder."
2. Claude analyzes your sent folder and generates a tone profile. For example, it might show an analysis like "You prefer short, clear sentences, begin greetings with 'Hello,' and close with 'Thank you, Gyeongzin Kim.'"
3. Now draft the replies. Enter the prompt as follows.
"For the emails classified as urgent and standard importance from earlier, use the my_email_tone.md file in this folder to draft replies that match my voice. Save each draft as a separate text file with the filename in the format draft_reply_[sender_name]_[date]."
4. Claude understands each email's context and drafts replies matching your tone profile. When email threads are long, Claude summarizes the full conversation context before drafting, so it won't repeat information already given.
You read the draft, adjust facts or tone, then send it directly from Gmail. Reply time drops from ten minutes per email to two or three minutes.
c Real-World Application: Automating Your Morning Email Briefing
When you turn your daily email sorting into a scheduled task, you arrive at work with a briefing document already waiting.
[Follow Along]
1. Type /schedule in the Cowork input field.
2. Configure the schedule settings as follows. Frequency: daily on weekdays (Monday to Friday). Time: 7:30 AM. Task: "Check unread emails in my Gmail and categorize them into three levels: urgent, standard, and reference. For urgent and standard emails, draft replies using my_email_tone.md. Save the categorization and drafts to the email_briefing folder with today's date in the filename."
3. Once saved, the task appears in the "Scheduled" section in the sidebar. From then on, Claude runs automatically at 7:30 AM each day.
The effect is greatest with regularly recurring email types: lecture requests, manuscript inquiries, meeting scheduling, quote requests. The details of each request differ, but the reply structure stays similar. Learning from past reply examples, Claude drafts responses to new lecture requests with available dates, fee rates, and needed information, and replies to manuscript requests with possible topics, length, and deadline details.
Remember one principle. A draft reply Claude creates must always be reviewed by a person before sending. In areas where one sentence matters,customer complaints, contract negotiations, legal correspondence,a factual error can have serious consequences. Claude takes on the labor of writing sentences, but the weight of the promises and responsibility in those sentences falls on the person who hits send.
d If You Run Into These Issues
(1) I connected the Gmail connector but got the message "Authentication failed." If two-step authentication is enabled on your Google account, you may need additional verification during the authentication process. Try reconnecting the connector after signing into your Google account in a browser first. If you're using an organizational account (Google Workspace), the administrator may need to allow connector access.
(2) The tone of the draft reply doesn't match my voice. If too few emails are analyzed when creating a tone profile, pattern recognition is inaccurate. Make sure to analyze at least twenty or more emails, and add a condition like "analyze only work emails" to avoid mixing business and personal messages. Open the generated tone profile file and correct any errors, and it will be more accurate next time.
(3) The scheduled task did not run. Scheduled tasks only execute when your computer is on and the Claude desktop app is open. If your laptop enters Sleep mode, it won't work. To have it automatically run before you
arrive at work, you need to set your computer's wake time ahead of the scheduled task execution time.
3 Calendar Integration and Meeting Preparation Automation
You have your first meeting at 9 a.m., but you do not check the calendar until 8:50 a.m. You remember the person's name, but their company affiliation is hazy, and you can barely recall what you discussed in the last meeting. In five minutes, you scramble through past emails and search for notes, but you cannot shake the feeling of walking in unprepared. On days when you have three or four meetings scheduled, the transition time between each meeting is essentially zero.
When you connect Claude to Google Calendar, this situation changes. Every morning, Claude analyzes today's and tomorrow's schedules and creates a briefing note for each meeting that includes attendee information, meeting purpose, and advance preparation items.
A Basic Use: Getting a Summary of Today's Schedule
Start by connecting the Google Calendar connector.
① Go to Settings in the Claude desktop app and navigate to Connectors.
② Find Google Calendar, click "Connect," and log in with your Google account.
③ After connecting, enter the following prompt in the Cowork tab:
"Check my Google Calendar for today's and tomorrow's schedules and organize them in chronological order. For each event, create a table with the meeting name, time, attendees, and location (or video conference link)."
④ Claude reads the calendar data and outputs a schedule summary.
What matters here is categorization, not just listing. Not all events carry equal weight. When you have a mix of external client meetings, internal decision-making sessions, simple check-in calls, and meetings that require travel, a summary like "10 a.m. Meeting A, 11 a.m. Meeting B" is not particularly helpful. You can add one line to the prompt: "Distinguish between external and internal meetings, and mark meetings that need advance preparation with an asterisk."
B Application Case: Auto-Generating Pre-Meeting Briefing Documents
Taking a step further from the schedule summary is automatically creating a preparation document for each meeting.
If the attendee is an external person, understand their company affiliation, title, and recent activities. If it is an internal meeting, summarize the current status of the related project.
① Create a "Meeting Prep" folder on your desktop, and if you have past meeting notes or related documents, place them in this folder.
② In Cowork, designate the "Meeting Prep" folder as your working folder and enter the following prompt:
"Check my Google Calendar for external meetings scheduled for today. For each meeting, create a briefing document in Word format with the following content.
First, provide an overview of the other party's company and recent news. Use web search to find articles from the past 30 days. Second, summarize any previous discussions with that company if notes are available in this folder. Third, draft an agenda for today's meeting. Organize it in the order of 'meeting purpose-discussion items-decisions needed-questions list.'
Name the file in the format 'Briefing_CompanyName_Date' and save it in this folder."
③ Claude extracts attendee information from the calendar, collects company information through web search, and references past notes in the local folder to generate the briefing document. Since this task combines multiple data sources, you will see sub-agents working simultaneously.
④ After completion, check the briefing document in the folder.
As this approach accumulates, you can also gain insights into your work patterns. By analyzing your calendar for a month, the data clearly shows what percentage of your work time is spent in meetings, which types of meetings are most common, and whether meetings tend to cluster on certain days.
C Real-World Application: Receiving Daily Morning Meeting Briefings Automatically
This task can also be turned into a scheduled workflow.
① Enter /schedule and configure it as follows. Execution frequency: run on weekdays Time: 7 a.m. Task content:
"Check all events on my Google Calendar for today and categorize them into external and internal meetings. For external meetings, create a one-page briefing document that includes the other party's company information and a draft agenda. For internal meetings, organize the current status of related projects and any carried-over items from the previous meeting. Save the complete results in the 'Meeting Prep' folder with today's date as the filename."
② When you arrive at the office each morning, today's briefing document is already in your "Meeting Prep" folder. By reading this document in the short time it takes to walk to the conference room, you can be as prepared as if you had spent days researching.
However, data has its limits. The briefing document shows the other party's recent financial performance, but the fatigue of internal conflict may fill the face of the person sitting across the table.
A machine presents conversation scenarios by combining facts and statistics, but it cannot tell you why the other person's eyes are wavering at this very moment. The machine has taken over the labor of understanding facts through documents, but the flexibility to read the atmosphere and change the direction of conversation in the field depends on human intuition.
D If You Encounter These Issues
(1) The calendar is retrieving the events, but attendee information shows only email addresses without names. This happens when the attendee's full name is not entered in the Google Calendar event. If you record the attendee's full name and company affiliation in the notes section when registering an event, Claude can create more accurate briefings.
(2) The company information found through web search is about a different company with the same name. Errors decrease if you include the company's exact English name, location, and industry in the prompt. Specify it clearly, like "ABC, an IT services company based in Gangnam-gu, Seoul."
(3) Personal events (doctor's appointments, family plans) are mixed in with your calendar and appear in the briefing. Add a condition to the prompt: "Focus only on business-related meetings. Exclude events that contain keywords like 'personal', 'doctor', or 'family.'" If you separate personal events into a separate calendar in Google Calendar, filtering becomes much cleaner.
4 Business Trip Preparation and Schedule Planning
A business trip is an extension of work, but the preparation process drains a different kind of energy. Searching for flights and accommodations, checking local transportation, calculating travel time between meeting locations, finding dining options, and estimating expenses are each small tasks, but together they add up.
For international business trips, you must also consider time differences, visas, currency exchange, and local communication options. Few people organize all of this into a single document. Most people head to the airport carrying fragments of information scattered through their minds.
When you hand trip-related files to Cowork, it gathers the scattered information and organizes it into three documents: an itinerary, an expense report, and a preparation checklist.
A Basic Use: Creating a Business Trip Itinerary
This is the most basic form. You provide the trip destination, duration, and main meeting schedule in text, and receive a day-by-day itinerary.
① Create a folder called "Tokyo Trip" on your desktop.
② In Cowork, select this folder and enter the following prompt:
"Prepare for a business trip to Japan, Tokyo from Monday through Wednesday next week.
Depart Incheon Airport Monday at 2 p.m. (flight not yet confirmed), meeting with Company A in Shibuya Tuesday at 10 a.m., meeting with Company B in Marunouchi at 3 p.m., depart Narita Airport Wednesday at 5 p.m. on return flight.
Based on this information, create a day-by-day itinerary in a Word document. Include transportation modes between each appointment (train, taxi) and estimated travel time. Recommend a hotel location convenient to both meeting venues.
Save with the filename 'Tokyo_Business_Trip_Itinerary' in this folder.
Claude conducts a web search to identify Tokyo's transportation information and hotel areas, then organizes the itinerary by time slot.
The most common error in itineraries is not the meeting times themselves, but the travel time between meetings. Claude calculates transit time by mode between two points and presents options like "Train from Shibuya to Marunouchi: 25 minutes, taxi: 35 minutes (varies by conditions)". An itinerary prioritizes arrival times over romance.
B: Application Case: Adding an Expense Forecast Report and Preparation Checklist
Create an itinerary that includes expenses and preparation items all at once.
[Follow Along]
In the same 'Tokyo Business Trip' folder, add any booking confirmation files you already have. Format doesn't matter: screenshots of flight confirmation emails, hotel voucher PDFs, etc. Cowork can read the contents of image files.
Enter the following prompt:
"Read all the booking confirmations in this folder and combine them with the itinerary you created earlier to produce the following two additional files.
First, create an expense forecast report in Excel. Divide the items into airfare, accommodation, transportation, meals, and miscellaneous, and show estimated amounts in both Korean won and Japanese yen. Search for today's exchange rate. Include subtotal and total formulas.
Second, create a preparation checklist in a Word document. Search for Tokyo's weather that week and include clothing advice. Separate business meeting items (business cards, presentation materials, etc.) from standard international travel items (passport, charging adapter, mobile Wi-Fi, etc.). Save both files to this folder."
Claude reads flight information from the screenshot, extracts check-in times from the hotel voucher, verifies exchange rates and weather through web search, then compiles all three documents.
In organizations requiring pre-approval for business travel, the expense forecast report can be attached as-is. After the trip, it serves as a benchmark document to compare projected versus actual costs. Small recurring items like airport rail, local taxis, gratuities, and meeting room beverages often emerge during post-trip settlement; Claude anticipates these details and preemptively includes them in the "miscellaneous" category.
C: Real-world Application: Assembling a Team Business Trip Package All at Once
Now imagine five team members traveling together instead of one person. Flight booking emails, hotel vouchers, and messenger screenshots shared with colleagues are all mixed together.
[Follow Along]
Add all booking-related files to the 'Tokyo Business Trip' folder regardless of format. PDFs, screenshot images, and text files can be mixed together.
Enter the following prompt: "This folder contains all booking confirmations for our five-person team's Japan business trip. Files are a mix of PDFs, images, and text. Read all these materials and create the following travel package.
First, a daily business trip itinerary (Word) visible to all team members. Organize flight numbers, airport arrival times, hotel check-in times, meeting times, and meal times by time slot. Include the mode of transportation and duration for each leg.
Second, a team-wide expense forecast report (Excel). Separate per-person costs from total team costs.
Third, a shared preparation checklist (Word). Organize weather, business etiquette (including notes on Japanese business card exchange protocol and gift-giving culture), and essential items. Save all three files to this folder."
Claude reads the contents of all files regardless of filename or extension. It extracts flight numbers and layover times from screenshots, pulls check-in times from hotel vouchers, and reorders everything chronologically.
Once this package is complete, repeated questions in the team group chat ("What time is the flight?", "What was the hotel name again?") disappear. You only need to share the itinerary file. Even in locations with unstable internet, the team can check the schedule using a pre-saved Word document.
For frequent business travelers, this structure itself becomes an asset. Change only the destination and dates, and a new travel document is quickly completed. Save it as a skill for even faster reuse: "Save the business trip preparation workflow I just ran as a skill named 'International Business Trip Package'."
For the next trip, one line suffices: "Apply the 'International Business Trip Package' skill to this folder."
Yet even the tightest machine-made itinerary can crumble before reality's whims. Airport baggage system failures, unannounced train strikes, and sudden road accidents exist beyond algorithmic foresight. Cowork eases the burden of trip preparation, but when lost in an unfamiliar city and needing to find a new route, that work falls to the person on the ground.
D: If These Problems Arise
(1) Text recognition from screenshot images fails. Low image resolution or small text reduces accuracy. Use PDF or text files when possible; if screenshots are necessary, capture them enlarged so text appears clearly.
(2) Calculated travel times differ from reality. Travel times Claude obtains through web search are average estimates. They don't account for rush-hour congestion, local construction, or detours caused by events. It's safer to include a note in the prompt to add roughly 30 minutes of buffer time to the itinerary.
(3) Exchange rates in the expense report become outdated. The rates Claude finds reflect the search date. They may differ from actual rates on the travel date, so add to your prompt: ask for a note in the report stating "Exchange rate applied as of search date; actual rates may vary".
Kim Kyung-jin
Attorney · Former Member of the National Assembly · AI Policy Researcher
© 2026 Kim Kyung-jin. All rights reserved.
















