On the conference room whiteboard, the distribution schedule by channel is drawn like a spider web. Blog, newsletter, landing page, email, presentation materials. Each channel requires different grammar and rhythm, and staff members explain the same product in different tones. While a copywriter squeezes out copy, a web developer translates that copy into code, and the sales team creates yet another version of presentation materials. The process that required several people to spend days bringing one new product to market takes on a different landscape in front of Cowork.

The six tasks covered in this chapter form the backbone of marketing practice. Building integrated campaigns, converting long-form content by channel, content gap analysis and calendar generation, B2B lead prospecting and scoring, SEO audits and website analysis, and creating customer onboarding packages. Each appears to be independent work, but they are tied together by one common point: people plan, Cowork executes, and people validate.

1 Building Integrated Marketing Campaigns

The afternoon landscape of the marketing department one month before a new product launch typically looks like this.

When the product manager hands over the functional specification, someone writes a blog post, someone else structures an email sequence, and another person creates presentation materials. The results that emerge from each person's corner address the same product yet have different tones. The phrase "smart work assistant" written in the blog becomes "next-generation automation platform" on the landing page.

This is because even when a brand guidelines document exists, each practitioner interprets it differently.

Cowork unites this fragmentation into a single flow. The core is to have one foundational document that readers can reference, and from that document, have blog posts, newsletters, HTML landing pages, email sequences, a 30-day launch plan, and PowerPoint presentation materials branch out.

There may be multiple channels, but the campaign should be one, and however many files there are, the message should flow as a single thread.

a Why It's Needed

Creating campaign content separately for each channel creates two problems.

One is time. If a copywriter, designer, web developer, and email manager each do their work, it usually takes more than two weeks.

The other is consistency. When blogs are educational, landing pages are sales-oriented, and emails feel urgent, customers wonder if these are messages from the same company.

It's more serious for solo entrepreneurs or small teams of three to five people. Without the resources to place professional staff in each channel, one person must handle all channels. When that one person gets tired, the quality of the entire campaign collapses.

b What It Does

When you hand a product guidelines document to Cowork, it extracts the core value proposition, target customer personas, and tone and manner from that document.

Based on this information, multiple sub-agents (auxiliary agents that perform individual tasks under instructions from the main agent) operate simultaneously.

Agent 1 writes the blog, Agent 2 codes the landing page HTML, Agent 3 structures the email sequence, and Agent 4 assembles the PowerPoint slides. They work independently while referencing the same guidelines document, so message consistency is maintained.

[To Know] What Are Sub-agents?

When you give Cowork a complex task, the main agent creates a work plan and then runs multiple auxiliary agents simultaneously. The progress of each sub-agent is displayed in real time in the right sidebar of the Cowork screen. It's like a team lead assigning work to team members who work at the same time.

c How to Do It: Basic Usage

Start with the simplest form. Let's create a blog post and an email at the same time.

Follow Along

① Create a folder called "New_Product_Campaign" on your desktop.

② Place the product specification (PDF or Word) inside it. Any document containing the product name, key features, target customers, and price will do.

③ Open the Claude Desktop app and click the "Cowork" tab at the top.

④ Set folder permissions in the left sidebar. Select the "New_Product_Campaign" folder to allow Claude to read and write files in this folder.

⑤ Paste the following prompt into the input field and press Enter.

"Read the product specification in this folder and create two pieces of content for me.

A blog post explaining the core value of the product (Word file, approximately 1,500 characters) An email announcement of the new product to send to existing customers (Word file, approximately 500 characters) Save both files in this folder."

⑥ Claude displays the work plan on the screen. Review the plan and click "Approve" or "Allow".

⑦ After 2-3 minutes, open the folder. The files Blog_Draft.docx and Email_Draft.docx have been created.

d How to Do It: Advanced Cases

Take a step further from basic usage. Add a brand guidelines document and increase the number of channels.

① Add the brand guidelines document to the "New_Product_Campaign" folder. It's a document containing the company's tone and manner, prohibited expressions, and frequently used phrases. If you don't have one, write the following content in a Word file and save it as "Brand_Guidelines.docx".

"Our brand's tone: professional yet approachable. Prohibited expressions: 'innovative,' 'groundbreaking,' 'first in the industry.' Preferred expressions: narratives centered on specific numbers and customer cases. Target customers: CEOs and executives of small to mid-sized companies with 50-200 employees."

② Also create and add a target customer persona document. One to two pages describing the customer's age, job title, concerns, and information consumption channels are sufficient.

③ Enter the following prompt in Cowork:

"Read all the product specification, brand guidelines, and customer persona documents in this folder. Based on the content of the three documents, create the following four pieces of content simultaneously.

A blog post with a subheading structure to rank well in search engines (Word file) A newsletter draft to send to subscribers (Word file), written from a different angle so it doesn't overlap with the blog A seven-email sequence for sequential sending over seven days (all 7 emails in one Word file), with day 1 addressing problem awareness, day 3 hinting at solutions, day 5 introducing the product, and day 7 driving action An internal executive summary of the launch strategy (Word file, 2 pages or less) Create a subfolder called 'Deliverables' and save all four files there."

④ Claude assigns sub-agents to work in parallel, with the progress displayed in the right sidebar.

⑤ When the work is complete, four Word files are generated in the "Deliverables" subfolder.

One important point to note: With email sequences, Cowork's draft proposes message structure and sending intervals, but when you transfer them to an actual sending system (Mailchimp, HubSpot, etc.), you must personally verify settings that affect sender reputation and delivery rate. Email is as important in the design of sending conditions as it is in message quality.

C. How to do it: Practical implementation

In practice, you request six outputs at once, including an HTML landing page and PowerPoint presentation.

1. In the folder, place the product specification document, brand guidelines, customer persona, along with the company logo image file and sample past marketing copy.

2. Enter the following prompt.

"I'm a marketing director. Please read all the documents in this folder. Based on this information, generate the following six integrated marketing packages at once.

Three blog posts covering the product's core value from different angles (each as a separate Word file), a draft new product announcement newsletter for existing customers (Word file), a single-page HTML landing page to collect pre-orders (HTML file) with hero section, feature showcase, customer testimonial area, and CTA button, a seven-day email drip sequence (Word file) with two subject line A/B test variations for each email, a launch strategy PowerPoint for internal executive reporting (PowerPoint file, around 10 slides), and a day-by-day 30-day launch marketing plan from D-30 to D-Day (Excel file) including channel distribution schedule, internal review dates, and ad budget spending timeline. Save all deliverables to a subfolder named 'Deliverables'."

3. Cowork launches six sub-agents simultaneously. PowerPoint skills, Excel authoring skills, and HTML coding capabilities are all deployed. The work typically takes 5 to 10 minutes.

4. Open and review each completed file one by one.

When reviewing, check whether the three blog posts address different angles, whether the newsletter and blog tones align with the brand guidelines, whether the landing page CTA language matches the final call-to-action in the email sequence, and whether the PowerPoint's logical structure aligns with the 30-day plan schedule. Once Cowork completes 90 percent, you take on the editor's role for the remaining 10 percent, adding the brand's unique warmth and strategic judgment.

D. Good prompts and bad prompts

Bad prompt: "Make some marketing materials for me."

This instruction reveals nothing about what product, who the audience is, which channels to use, or what file format to request. Cowork will produce something even without context, but the result will be nearly useless.

Good prompt: "Read the product specification and brand guidelines in this folder, then create one blog post (Word, 1,500 characters) targeting executives at mid-sized companies with 100 employees and a seven-day email sequence (Word, Day 1 problem awareness / Day 3 case study / Day 5 product introduction / Day 7 call-to-action) and save them to the Deliverables folder."

A good prompt contains four elements: reference documents, target audience, output format and length, and storage location. The more clearly you specify these four elements, the higher the quality of the output.

E. If these issues arise

(1) The blog and newsletter tones are too similar. Cause: You did not specify channel-specific tone differences in the brand guidelines document. Add this distinction to the guidelines: "The blog takes an educational, in-depth tone; the newsletter uses a short, curiosity-driven tone." Then submit the request again.

(2) The HTML landing page appears broken in the browser. Cause: The HTML Cowork generates has basic structure and text focus. CSS styling may be insufficient. Either pass it to a web developer or give Cowork the additional instruction: "Apply mobile responsive design."

(3) The company logo did not appear on the PowerPoint slides. Cause: Even if you place the logo file in the folder, Cowork may not automatically insert it unless you explicitly state in the prompt: "Place the company logo in the bottom right corner of each slide." Supplement your instructions with this specificity.

2. Multi-Channel Repurposing of Long-Form Content

It took two days to shoot and edit a 40-minute YouTube lecture. The content is dense. Yet only 300 people watched the video.

Post the same content on LinkedIn and 5,000 people could see it; send it as a newsletter and reach 2,000 subscribers. The problem is that adapting a 40-minute script to fit each platform's conventions takes as much time as producing the video itself.

A. Why it's necessary

The strategy of creating multiple channel versions from a single original piece of content is commonly called "One-Source Multi-Use."

The strategy has existed for a long time, but execution has always been the bottleneck. To convert a YouTube script into a LinkedIn post, you must extract the core idea, rewrite the opening sentence for the platform, and add hashtags.

Converting the same material into a Twitter(X) thread requires breaking your thoughts into 280-character chunks. At one to two hours per channel, five channels eat up an entire day.

B. What it does

When you hand Cowork a long script or transcript, it extracts the message suited to each channel. LinkedIn posts place industry insights upfront because professionals read them there; Twitter threads develop arguments in short bursts; blog outlines use subheading structures matched to search intent; emails deliver a single core message.

Here lies the most critical task: infusing brand voice,your brand's distinctive tone and manner of speaking. To strip away that characteristic smooth, arid smell of AI-generated text, you must include samples from your own past writing.

Cowork learns your cadence, sentence length, and recurring expression patterns from those samples.

C. How to do it: Basic use

Start by converting a single YouTube script into one LinkedIn post.

Follow along: 1. Create a folder named "Content_Repurposing."

2. Place the YouTube video's script text file (or caption file) in the folder. To download captions from YouTube, click "Show transcript" in the menu at the bottom of the video, then copy the text and save it as a text file.

3. Connect this folder in Cowork and enter the following prompt.

"Read the YouTube script in this folder and create one LinkedIn post. Place the most counterintuitive claim or data from the script at the start to grab attention in the opening sentence. Write in an expert tone, but end with a question that prompts readers to share their experience and start a conversation. Save it as a Word file."

4. Claude analyzes the script and generates a LinkedIn post draft as a Word file.

D. How to do it: Advanced application

Expand to three channels and add a brand voice file.

1. Add a file called "Brand_Voice.txt" to the folder. Write it like this: "My speaking style: I alternate between short and long sentences. I use metaphors often but never overstate. I don't use emoji. Phrases to avoid: hedging endings like 'could be said to' or 'amounts to.' Preferred closings: end with a question or a single declarative sentence."

2. Also place two or three LinkedIn posts or emails you wrote yourself in the folder as text files. These become the learning material from which Cowork picks up your voice.

3. Enter the following prompt.

"First read the 'Brand_Voice.txt' file and my past writing samples in the folder, then familiarize yourself with my voice and phrases to avoid. Next, read the YouTube script and convert it into the following three pieces of content.

LinkedIn post (expert tone, key insights focused, hook in first line) X thread 5-7 posts (each tweet reads independently and makes sense, with original video link space in final tweet) Email newsletter (for subscribers in a 'here's this week's highlights' tone, sharing 2-3 key points only) Three contents should each cover different insights as the main focus. Save all individually as Word files.

④ Cowork first analyzes the brand voice file, then runs three sub-agents simultaneously to generate content for each channel.

D. How to do it: practical application

In practice, expand to 5 channels and save this workflow as a Skill (a reusable set of work rules saved for future use) so the next time you can run it with a single command.

① With script file and brand voice file ready in a folder, enter the following prompt.

"Read the YouTube script and repurpose it into the following five content types. LinkedIn carousel post (structure three key insights as slides) X thread (5-7 threads starting with a hook) Email newsletter (storytelling format, 2-3 key points) Three YouTube Shorts scripts (60 seconds each, extract interesting parts only, include timestamps) Blog post outline with SEO keywords (include H2 and H3 tag structure) Ensure each content covers different insights as the main focus without overlap. Save all individually as Word files."

② If satisfied with the results, save this workflow permanently. "Save the entire content conversion process you just performed as a Skill named 'YouTube-to-Social'."

③ Next time you put a new script in the folder and enter /YouTube-to-Social, Cowork will automatically generate content for five channels following the saved rules.

E. It is worth making conversion rule tables first. When repurposing content across multiple channels, setting channel-specific conversion rules in advance stabilizes quality. Here is an example.

LinkedIn: break into single sentences, put real-world problems up front. 3-5 hashtags. X: short and concise. One claim per tweet. First tweet is a hook, last is a CTA. Blog: establish context before argument. Headline structure matching search intent. Email: request only one action. Put the core message in the subject line.

If you keep this rule table in a text file in the folder, Cowork references it each time.

F. If this problem occurs

(1) Channel outputs look like they're simply summarizing the original script. Cause: the prompt is missing the instruction to focus on different insights for each channel, or you haven't clearly distinguished each channel's role. Specify angles like: LinkedIn from an industry trend analysis perspective, X focusing on counterintuitive facts, email emphasizing personal experience narratives.

(2) Brand voice wasn't reflected. Cause: either the past writing samples are too few, or the brand voice file instructions are vague. The more specific you write,keep sentences under 40 characters, draw metaphors from everyday objects,instead of prefers short sentences, the more effective it is.

3. Content gap analysis and calendar generation

Assume you've been running a blog for two years and written 150 posts. Each time you plan new content, you wonder: have I covered this topic before? When you search, you find three posts with similar titles, yet there's not a single post addressing questions your readers ask most frequently.

A. Why it's needed

Two problems grow as content accumulates: duplication and gaps. When you repeat the same topic with only slight variations, readers feel déjà vu and traffic plateaus. Simultaneously, if posts addressing market-driven questions are missing, competitors claim that ground first. A person would need over a day to read all 150 past posts and identify patterns, but Cowork delivers structured analysis instantly upon receiving a list.

B. What it does

When you give Cowork past content files, it classifies everything by topic and analyzes which areas are overproduced and which are blank. For instance, in an AI-focused blog, generative AI tool reviews might have 12 posts while AI ethics and regulation has only 2. Once gaps are identified, it derives new content ideas that don't overlap with existing posts and completes them as a 3-month calendar placed on a timeline.

[Know this] What is content gap?

Content gap divides into four types.

Subject gap is a question you haven't covered at all. Format gap is when you cover the same topic in blog form only and haven't adapted it to short social formats. Audience gap is when you have only beginner posts and no decision-maker materials. Funnel gap (funnel refers to the stages customers pass through from awareness to purchase) is when you have many awareness-stage posts but comparison and purchase-stage posts are absent.

C. How to do it: basic usage

You can do basic analysis with just past blog title lists.

Try it. ① Create a folder called Content_Analysis.

② Organize past blog post titles and URLs in Excel or CSV format. Four columns suffice: title, publication date, category, URL.

③ Put this file in the folder, connect it in Cowork, then enter the following prompt.

"Read the blog list file in this folder and classify the titles by topic. Identify which topics have the most posts and which have the fewest, then organize the top 5 of each and save as a Word file."

④ Claude generates the classification result as a Word file. This alone shows at a glance what we've talked about most and what we haven't.

D. How to do it: application examples

When you use actual blog original files instead of title lists, the depth of analysis changes.

① Put 20-50 past blog articles' full texts in your folder as text files or PDFs. You can also add podcast episode scripts the same way.

② Enter the next prompt.

"Analyze all past content files in this folder. Extract the top 5 themes and keywords I've covered most frequently. Based on past data, derive 20 new content ideas I've never covered before (avoiding duplication). Each idea should include a concrete title and a 3-line summary.

Save results as a Word file."

③ Cowork analyzes text patterns, finding repeatedly appearing keywords and identifying blank areas yet uncovered.

There is one point to note here. The blank space Cowork finds isn't necessarily a space you should fill.

Why a certain topic wasn't covered in the past may be because it didn't align with our brand identity or it was a sensitive area we strategically chose to avoid.

Finding statistical gaps is the machine's job, but determining whether filling that gap is strategically sound is yours.

E. How to do it: practical application

Arrange the 20 ideas into a 3-month calendar.

① With 20 idea files generated from application cases in your folder, enter the following prompt:

"Create a three-month content calendar based on the 20 content ideas derived above. The conditions are as follows: publication twice a week; alternate search-type, case-study-type, comparison-type, and conversion-type pieces to avoid topic concentration; alternate between data-research-heavy deep analysis pieces and lighter experience-based pieces to spread out creator energy expenditure; in each row, include publication date, content title, the one-line question this piece answers, and distribution channels (blog/LinkedIn/newsletter). Save it as an Excel file."

② Claude generates an Excel file. Once you share this calendar with team members, it becomes a collaboration tool you can use right away.

There is a reason for including "the one-line question this piece answers" in the calendar. Even if the same topic is covered, it is not a duplicate if the question differs; even if the title changes, it is a duplicate if the reader's question is the same. With this one line, you can immediately see whether similar pieces are appearing back-to-back.

If You Encounter These Problems

(1) The 20 derived ideas are too abstract. Cause: You included only a list of past content titles, not the full texts. Claude cannot discern the depth and angle of the writing from titles alone, so include the original text files when possible.

(2) The three-month calendar distribution is unrealistic. Cause: Claude cannot know the exact production difficulty of your writing. Adding realistic conditions to the prompt such as "data-analysis pieces take 5 days to produce, experience pieces take 2 days" will improve the distribution.

4 B2B Lead Generation and Scoring

Your sales team has a CSV file with prospects from 1,000 companies. How many of them are likely to actually buy your product? If a salesperson reviews all 1,000 one by one, it takes two weeks.

Why It Matters

The core of B2B sales is deciding who to meet first. You have many names but little time. The difference between sending identical emails to 1,000 companies and sending customized emails to your top 50 shows clearly in reply rates. Lead scoring,assigning scores to prospects based on fit and interest,is the work of organizing these priorities into numbers.

What It Does

Give Claude a CSV file of prospects and your Ideal Customer Profile (ICP) criteria, and it scores each company against those criteria. It identifies decision-makers at high-scoring companies and drafts customized cold emails tailored to each company's situation.

How to Use It: Basic

First, learn the principles of scoring with a small list.

Follow along: ① Create a "B2B_Sales" folder.

② Place a CSV file with information on 30 prospects inside it. The columns are: company name, industry, employee count, annual revenue, contact name, job title, email.

③ In the same folder, create a file called "ICP.txt" and write your ideal customer criteria.

"Our ideal customer: SaaS or IT services, 50-500 employees, 5-50 billion won annual revenue, bonus points for fundraising in the past two years."

④ In Claude, enter the following prompt:

"Analyze the 30 companies in this folder's CSV file against the ICP.txt criteria. Score each company from 1 to 100 points and explain your scoring by category. Save the results to an Excel file. Format the columns as: company name, score, scoring rationale, industry match, size match."

⑤ Claude generates an Excel file. When you open it, you see scores and rationales organized like this: "Company A: 82 points. Industry match (+25), employee count in range (+20), revenue meets criteria (+20), fundraising history unconfirmed (+0), other (+17)."

How to Use It: Advanced

Generate customized cold email drafts for high-scoring companies.

① With the Excel file from the basic section in your folder, enter the following prompt:

"From the lead scoring results I just generated, filter for companies with 70 points or higher. For each company, do the following: check the contact's job title from the CSV and mark the decision-maker (CEO, CTO, VP level).

Draft a cold email for each company. In the opening sentence, mention something specific about their industry or size. In the second paragraph, explain what value our service can bring. End by proposing a 15-minute call. Generate a master Excel file containing the lead score, decision-maker info, and draft email for each company."

② Claude fills in a separate column in the Excel file with cold email drafts, each with a different opening sentence tailored to the company.

Adjusting Weights and Validation

Your sales team may disagree with the scores. If you decide that annual revenue matters more than employee count, add weights to your ICP.txt.

"Apply a 2x weight to the revenue item and 1x weight to the employee count item." Then run it again.

It is best when sales and marketing agree on the scoring criteria. Numbers look objective, but the standards that produce them are human choices. Without agreement on standards, scores will not survive long in sales meetings.

If You Encounter These Problems

(1) Scores come out similar for all companies. Cause: Your ICP criteria are too loose or have too few items. Just "industry match" and "size match" will not differentiate enough. Add specific criteria like "recent job postings include IT roles" or "company news in the past year mentions digital transformation."

(2) The cold emails are too generic. Cause: Your CSV lacks company-specific details. Beyond company name and industry, add columns like "recent news," "core services," and "competitors." This raises the personalization level of the emails.

(3) Claude cannot search for company information on the web. Cause: The domain is not allowed in Claude's network access settings, or Claude in Chrome browser integration is not activated. Check the network allowlist in settings.

5 SEO Audit and Website Analysis

You have been running a website and asking yourself, "Why does it not appear in search?" You post content steadily and the design is clean, yet your pages are buried after page 3 of Google results. The cause usually hides in technical details invisible to the eye.

Why It Matters

An SEO audit,systematically checking your search engine optimization status,is not about manipulating keywords. It is about clearing the path for search engines and users. Empty meta tags, missing alt text on images (hidden text describing an image), slow page load speed, and broken layouts on mobile screens all drag down search rankings. It is hard for a human eye to check all these items at once, and professional SEO consulting is expensive.

What It Does

Give Claude your website URL, and it retrieves and analyzes the HTML structure of that page.

The audit items fall into four main areas: on-page SEO (title tags, meta descriptions, H1-H6 hierarchy, keyword placement), technical SEO (loading speed, render-blocking elements, cache settings), content quality (body length, keyword density, duplication), and user experience (navigation intuitiveness, CTA placement, mobile responsiveness).

When Claude in Chrome is enabled, Cowork can open an actual browser, visit the page, and evaluate the visual layout as well.

How to Use It: Basic

Let's audit one main page of your company website.

Follow Along

① Create a folder called "SEO_Audit".

② Write one URL of the webpage to be audited in a text file and put it in the folder.

③ Enter the following prompt in Cowork.

"Analyze the webpage at the URL in this folder. Check the following items and save the results in a Word file.

Title tag length and keyword inclusion; existence and quality of meta descriptions; whether H1 tag appears only once and hierarchy is correct; whether images have alternative text (alt text); number of internal and external links. For each item, mark the status as 'Good', 'Needs improvement', or 'Needs immediate fix'."

④ Claude analyzes the HTML of the URL and generates the audit results in a Word file.

How to Use It: Applied Example

Audit multiple pages at once and add UX analysis.

① Write the URLs of your top five traffic pages in a CSV file and put it in the folder.

② Verify that the Claude in Chrome connector is enabled.

③ Enter the following prompt.

"Using the browser, visit the five webpages in this CSV directly. SEO audit: meta tags, heading structure, keyword placement, broken links, outdated information check. UX analysis: whether the value proposition is visible on the first screen, CTA button placement and visibility, mobile responsiveness. Classify each page into three levels: 'Needs immediate fix', 'Improvement recommended', 'Long-term task'. Generate a Word audit report organized by page."

④ Cowork opens a browser, visits each page (the tab border displays in orange), and creates a report by analyzing both text and visual layout together.

Suggested Report Structure

A good audit report divides into four bundles: indexing and crawling (robots, noindex, sitemap, internal links), on-page structure (title, H1, image ALT, duplicate meta), page experience (HTTPS, speed, mobile, layout stability), and conversion UX (CTA placement, form count, trust elements). This division lets the development team, content team, and design team each take their own tasks immediately. A report that says "fix everything" typically accomplishes nothing. A report that says "fix only these three things this week" actually changes the site.

If You Run Into These Issues

(1) Claude cannot access the URL. Cause: The domain is not in the allowlist in Cowork's network settings. In Claude Desktop app Settings, add the domain of the site you want to analyze to the Network Allowlist.

(2) UX analysis comes out as text only. Cause: Claude in Chrome is not enabled, so visual analysis could not be performed. In connector settings, verify that the Claude in Chrome extension is installed and connected.

6 Creating a Customer Onboarding Package

The 48 hours right after signing a contract with a new client set the tone for the relationship. If there is no contact for a week after signing the contract, the client begins to wonder, "Does this company really care about our project?"

Why It Matters

Client onboarding typically requires five documents: proposal, service agreement, invoice, project timeline, and welcome email. Each looks separate, but they are actually related documents derived from a single client record. When client name, service scope, amount, or schedule changes, all five documents must update at the same time.

Doing this manually usually takes an hour or more. You find and open a template, change names and amounts, convert to PDF, and attach to email,over and over. On busy days, you end up with the invoice amount different from the contract amount.

What It Does

When you give Cowork the client information, it generates six documents simultaneously: proposal (PPT), service agreement (Word), invoice (Excel), project timeline (Excel), welcome email (text file), and internal onboarding checklist (Word). If you have existing template files in the folder, Cowork reads the format and fills in the blanks.

How to Use It: Basic

Let's start simply by creating two things: a welcome email and a project timeline.

Follow Along① Create a folder called "2026_NewClients" and connect it to Cowork.

② Enter the following prompt.

"New client info. Company name: GlobalTech, Service: AI chatbot development, Total amount: 20 million won, Project duration: 4 months.

Create two documents based on this information.

A welcome email draft ready to send to the client (text file), including an outline of next steps, a request to schedule the first meeting, and contact information for the account manager;

A four-month weekly project timeline (Excel file) including kickoff, material submission, first review, revision, and delivery milestones. Save both files to this folder."

③ Two files are created in the folder after two to three minutes.

How to Use It: Applied Example

Increase the number of documents and use existing templates.

① Put the invoice format (Excel) or contract template (Word) you normally use in the folder. A blank template with your company logo is fine.

② Enter the following prompt.

"Using the invoice template and contract template in this folder as reference, fill in the blanks customized for the GlobalTech client. Additionally, create a new proposal (Word) and internal onboarding checklist (Word). In the proposal, organize the agreed-upon service scope, each party's roles, and performance measurement criteria. In the onboarding checklist, organize the tasks to be completed in the first week by item."

③ Cowork reads the structure of the existing templates, maintains the logo and formatting, and fills in only the customer information.

How to Do It: Practical Application

Generate six documents in parallel at once.

① Enter the following prompt.

"Create an onboarding package for the GlobalTech client.

Service: AI chatbot development, Total amount: 20 million won, Payment terms: 50 percent upfront at start / 50 percent upon completion, Project duration: 4 months. Generate six files simultaneously. Project proposal (PPT) with agreed service scope, each party's roles, and KPI specifications. Service contract (Word) with service level agreement (SLA), confidentiality clause, and termination conditions. 50 percent upfront invoice (Excel) with formula-based amount calculation. Monthly timeline by week (Excel) with key milestones and deliverables organized. Welcome email draft (text file) with project progress guidance, first meeting coordination, and contact information. Internal team onboarding checklist (Word) with tasks for the first two weeks. Save all files to the 'GlobalTech_onboarding' subfolder."

② You will see six sub-agents executing simultaneously in the right sidebar of Cowork.

③ After completion, open the folder and six files will be generated. Always verify that the upfront invoice amount is calculated correctly as 10 million won, the contract duration matches the timeline duration, and the kickoff date in the welcome email aligns with week one of the timeline.

If This Problem Occurs

(1) The invoice amount does not have a formula. Cause: If you don't explicitly state "formula-based amount calculation" in the prompt, Cowork may enter numbers directly. Give a specific instruction: "Insert a formula in the total amount cell that automatically calculates 50 percent."

(2) Information is inconsistent across the six documents. Cause: Sub-agents working in parallel may have interpreted information differently. Clearly declare common variables in the first line of your prompt: "Use the following common information consistently across all documents: Company name GlobalTech, Amount 20 million won, Duration 4 months, Start date April 1, 2026."

The six tasks covered in this chapter each have a distinct purpose.

Integrated campaign building unifies fragmented channel messages into one, content repurposing expands the reach of a single original piece, gap analysis shows what was said and what was not said, lead scoring picks the 50 prospects to meet first from 1,000 targets, SEO audit uncovers invisible technical bottlenecks, and onboarding package reduces administrative time immediately after contract signing.

Cowork's role across these six tasks is consistent. It performs repetitive, time-intensive execution in your place. The role people must play is also consistent. Set the direction of the strategy, add brand-specific judgment to the output, and maintain legal and ethical boundaries. When Cowork completes 90 percent, people work in the remaining 10 percent. That 10 percent determines whether the output is useful.