Claude Cowork — Hands-On Workshop

From data file to
boardroom-ready insight
in under 10 minutes.

Three guided walkthroughs and four independent challenges using realistic demo data. No setup required — just point Claude at a folder and work through the exercises at your own pace.

Format
3 Walkthroughs + 4 Challenges
Data Files
5 Demo CSVs
Company
Acme Corp (Fictional)
Category
Business Automation
🚀
Start Here

Getting Started with Cowork

Cowork is Claude working interactively with your files — reading, analysing, and building outputs in real time. This guide uses five fictional demo data files representing a mid-size manufacturing and distribution company called Acme Corp.

What is Cowork?

Cowork is a distinct mode within Claude that goes beyond conversation. You point it at files and folders — spreadsheets, CSVs, PDFs, documents — and Claude reads them in full, reasons across them together, and produces structured outputs: summaries, charts, memos, models. Think of it as giving Claude a desk and a brief, not just a question. It can cross-reference multiple files simultaneously, notice patterns you might miss, and format its findings however you need them.

⚠️
Requirement: Cowork is available on Claude Pro and Team plans. If you are on the free tier, you will not be able to access Cowork mode. Check your plan at claude.ai before starting.
How to start a Cowork session
1

Go to claude.ai and open Cowork mode

Log in to claude.ai. At the top centre of the screen, you will see a mode selector. Click Cowork to switch into the mode where Claude can work interactively with your files and produce richer outputs like charts and formatted documents. Do not use standard chat for these exercises — you need Cowork.

Look for this at the top of the screen:
Chat Cowork Code
2

Create your demo data folder

Create a folder on your computer called cowork-demo-data. Download all five CSV files from the section below and save them into this folder. You will point Cowork to this folder for every exercise.

Setup guide

1

Create a new folder on your computer

On your Desktop or in Documents, create a folder called:

📁 cowork-demo-data
2

Download the 5 demo files and save them into this folder

Use the download cards further down this page. Save each file into your cowork-demo-data folder.

📁 cowork-demo-data 5 files
📄 Q3-2026-Revenue-by-Division.csv
📄 Q3-2026-KPIs.csv
📄 Supplier-Variance-Report-Raw.csv
📄 Q3-2026-Cost-Analysis.csv
📄 Weekly-Market-Prices.csv
3

Point Cowork to this folder in your first prompt

Cowork reads everything in the folder automatically. You don't need to attach files one by one.

3

Point Cowork to your folder

In Cowork, tell Claude where your files are. You can reference the folder in your first prompt — for example: "I've put my demo data files in the cowork-demo-data folder." Cowork will read all the files in the folder automatically. You do not need to attach files individually.

4

Send your opening prompt

The exercises include exact prompts to copy. Paste the prompt and send. Claude will read all the files in your folder and respond with a structured analysis. You don't need to explain the file format — Claude figures it out.

5

Iterate with follow-up prompts

Each walkthrough has iteration steps — follow-up prompts that refine, reformat, or deepen the output. You can also try your own variations. The files stay available for the whole conversation.

🔄
Tip: You can work through all three guided walkthroughs in a single Cowork conversation. Claude retains context from earlier exercises, which means it already understands the Acme Corp data by the time you reach Walkthrough 3. For the independent challenges, consider starting a new conversation if you want a clean slate.
Demo data files for this workshop

All five files are fictional. Data is illustrative only and does not represent any real company, supplier, or market. All figures have been generated for training purposes.

Click any file card below to download it. Save all five files into your cowork-demo-data folder.

📊

Q3 Revenue by Division ⬇ Download CSV

Q3-2026-Revenue-by-Division.csv

Quarterly revenue figures across Acme Corp's four divisions (Acme Industrial, Acme Consumer, Acme Materials, International) broken down by product line. Includes actuals vs. budget and year-on-year comparison columns. Used in Walkthrough 1 and Challenges A, C and D.
📈

Q3 KPI Dashboard ⬇ Download CSV

Q3-2026-KPIs.csv

Key performance indicators for Q3 2026 including gross margin, EBITDA, headcount efficiency, volume throughput, and customer service metrics. Each KPI includes a target, actual, RAG status, and variance. Used in Walkthrough 1 and Challenges A and B.
🏭

Supplier Variance Report ⬇ Download CSV

Supplier-Variance-Report-Raw.csv

Procurement data for 15 key suppliers showing contract value vs. actual spend for Q3 2026. Includes supplier name, category, region, contract value (EUR), actual spend (EUR), variance (EUR and %), payment terms, and last review date. Used in Walkthrough 2 and Challenges A and B.
💰

Cost Analysis by Category ⬇ Download CSV

Q3-2026-Cost-Analysis.csv

Departmental and category-level cost breakdown for Q3 2026. Covers raw materials, packaging, energy, logistics, labour, and overheads. Budget vs. actual with variance flags. Used in Walkthrough 2 and Challenges A and C.
📉

Weekly Market Prices ⬇ Download CSV

Weekly-Market-Prices.csv

13-week rolling commodity price data for key input materials: Steel Grade A (EUR/t), Aluminium Grade B (EUR/t), Copper Standard (EUR/t), UK Steel Index (GBP/t), EU Metals Index (EUR/t), Industrial Resin (EUR/t), and Diesel (EUR/L). Used in Walkthrough 3 and Challenge D.
💡
Tip: For the independent challenges, all five files are already in your cowork-demo-data folder — Claude will cross-reference them automatically when you ask questions that span multiple datasets.
Ready when you are — start with Walkthrough 1
📊
Guided Walkthrough 1 of 3

Q3 Financial Snapshot

Turn two raw data files into a polished executive summary — headline revenue, year-on-year comparison, division breakdown, and KPI traffic lights — then iterate toward a CEO-ready narrative.

🔄
Make sure you are in Cowork mode and have pointed Claude to your cowork-demo-data folder. If you haven't already, open a new Cowork conversation and tell Claude where your files are.
Select Cowork at the top of claude.ai:
Chat Cowork Code
Files needed for this exercise
  • Q3-2026-Revenue-by-Division.csv
  • Q3-2026-KPIs.csv
Objective

Experience how Claude reads multiple data files simultaneously, connects them together, and produces a structured one-page summary — without you having to explain the data format or structure the output yourself.

Step 1
Opening prompt — generate the executive summary

Make sure both files are in your cowork-demo-data folder, then send the following prompt. Pay attention to how Claude structures its response — it will use the file contents without you needing to explain the column names.

Prompt to copy
I've pointed you at my cowork-demo-data folder. Use Q3-2026-Revenue-by-Division.csv and Q3-2026-KPIs.csv for Acme Corp. Please create a one-page executive financial summary for Q3 2026 covering: 1. Headline group revenue (Q3 2026 actual vs. Q3 2025, with year-on-year change) 2. Division breakdown — Acme Industrial, Acme Consumer, Acme Materials, International — each with Q3 revenue and year-on-year comparison 3. Top 3 KPIs with RAG (Red/Amber/Green) status 4. A two-sentence narrative on overall performance Format it cleanly with clear section headings. Use plain numbers — no need for charts yet.
What to look for in Claude's response
  • Does it correctly identify the four divisions and their revenues?
  • Does it calculate variances rather than just repeating raw numbers?
  • Does the RAG status match the KPI targets in the file?
  • Is the two-sentence narrative grounded in the data, not generic?
Step 2
Iteration — add a visual revenue chart
Follow-up prompt
Prompt to copy
Now add a visual bar chart showing Q3 revenue by division — Q3 2026 actual vs. Q3 2025. Render it as a proper chart I can view visually. I want to see the relative sizes at a glance and the year-on-year change for each division.
What to look for
  • Cowork can render visual charts and HTML outputs directly — you should see a proper chart, not just text.
  • Does the chart clearly differentiate Q3 2026 from Q3 2025?
  • Are the proportions between divisions visually accurate?
  • Try asking for a different chart type (e.g. "show this as a horizontal stacked bar chart") and compare the outputs.
Step 3
Iteration — CEO "so what" paragraph
Follow-up prompt
Prompt to copy
Write a "so what" paragraph for the CEO — two to three sentences that cut through the numbers and explain what the Q3 data actually means for the business. Flag the single most important thing that needs attention this quarter. Be direct, no corporate hedging.
What to look for
  • Is the paragraph grounded in the specific data from the files — or generic?
  • Does it identify the right "most important thing" based on the KPI statuses?
  • Notice how Claude's tone shifts when you explicitly ask for directness.
💡
Bonus iteration: Ask Claude to reformat the entire one-pager as a polished PDF you can download and share. Try asking for "a clean, professional layout with tables and colour-coded RAG indicators, formatted as a downloadable PDF."
🏭
Guided Walkthrough 2 of 3

Supplier Variance Drill-Down

Identify suppliers over contract, cross-reference to cost category budgets, and have Claude draft a procurement memo with recommended actions — all from two raw CSV files.

Files needed for this exercise
  • Supplier-Variance-Report-Raw.csv
  • Q3-2026-Cost-Analysis.csv
Objective

Practice using Claude as a procurement analyst — flagging contract breaches, connecting line-level spend data to category budgets, and producing a professional memo that a Head of Procurement could send without edits.

Step 1
Opening prompt — flag suppliers over contract
Prompt to copy
In my cowork-demo-data folder, use these two files: Supplier-Variance-Report-Raw.csv (supplier contract values and actual spend for Q3) and Q3-2026-Cost-Analysis.csv (cost category budgets). Please do the following: 1. Identify every supplier where actual spend is more than 5% above the contract value 2. For each flagged supplier: show supplier name, category, contract value, actual spend, variance %, and estimated annual impact if the Q3 overage rate continues for a full year 3. Cross-reference these flagged categories to the Q3 cost analysis — are these categories already over budget? 4. Rank the flagged suppliers by financial impact (highest first) Present this as a structured table.
What to look for in Claude's response
  • Does it correctly calculate variance % (not just flag based on absolute values)?
  • Does the category cross-reference make sense — are the same categories flagged in both files?
  • Is the ranking by financial impact correct?
  • Note any supplier where Claude flags additional context worth investigating.
Step 2
Iteration — draft a procurement memo
Follow-up prompt
Prompt to copy
Now draft a formal procurement memo based on this analysis. It should be addressed to the Head of Procurement from the Finance & Commercial Analytics Team. Structure: - Subject line - One-paragraph executive summary of the issue - Table of flagged suppliers (top 5 by impact) - Recommended actions for each supplier — one specific action per row (e.g. renegotiate, escalate, review contract terms) - A closing paragraph on the budget risk if no action is taken Keep it professional, specific, and under 400 words.
What to look for
  • Are the recommended actions supplier-specific, or generic filler?
  • Does the closing paragraph quantify the budget risk using the data from the files?
  • Could this memo be sent as-is, or does it need significant editing?
  • Try asking Claude to adjust the tone — more urgent, or more diplomatic — and compare.
Step 3
Iteration — add a supplier risk heat map
Follow-up prompt
Prompt to copy
Create a simple risk matrix for the flagged suppliers. Score each one on two axes: financial impact (low / medium / high) and ease of switching supplier (easy / moderate / difficult). Present this as a 3x3 text grid and explain the logic for each placement.
What to look for
  • Does Claude apply reasonable business logic to "ease of switching" even though it's not in the data?
  • Does it explain its assumptions transparently?
  • Is the risk matrix useful as a prioritisation tool?
💡
Try this: Ask Claude to produce the memo as a downloadable Word document. Cowork can generate .docx files directly — open it and see how well the formatting, tables, and structure come through.
📉
Guided Walkthrough 3 of 3

Market Price Trend Monitor

Turn 13 weeks of commodity price data into a trend analysis — identify materials moving outside normal range, chart the movements, and translate price shifts into margin implications for the business.

Files needed for this exercise
  • Weekly-Market-Prices.csv
Objective

Practise using Claude as a market intelligence analyst — scanning price series for anomalies, flagging risks, visualising trends, and producing a concise management briefing on input cost exposure.

Step 1
Opening prompt — trend analysis and anomaly detection
Prompt to copy
In my cowork-demo-data folder, use Weekly-Market-Prices.csv which contains 13 weeks of commodity price data for Acme Corp's key input materials. Please: 1. Calculate the 13-week average price for each commodity 2. Flag any commodity where the most recent week's price is more than 3% above or below the period average 3. For each flagged commodity: show the average, the latest price, the % deviation, and the direction (rising / falling) 4. Chart all 7 commodities as a visual trend chart showing price movement across the 13 weeks 5. Write a two-paragraph summary of the key price movements and what they suggest about input cost pressure for Q4 Commodities in the file: Steel Grade A (EUR/t), Aluminium Grade B (EUR/t), Copper Standard (EUR/t), UK Steel Index (GBP/t), EU Metals Index (EUR/t), Industrial Resin (EUR/t), Diesel (EUR/L).
What to look for in Claude's response
  • Are the 13-week averages calculated correctly across all weeks (not just selected rows)?
  • Does the 3% threshold flag the right commodities?
  • Is the trend chart readable — can you see peaks and troughs clearly?
  • Does the summary correctly connect price movements to business implications (not just describe the numbers)?
Step 2
Iteration — margin impact analysis
Follow-up prompt
Prompt to copy
Using the commodity price report already created, add a new Section 5 — Margin Impact Analysis — to the bottom of that document, before the footer. For the commodities flagged as rising more than 3% above average, estimate the margin impact for Acme Corp. Assume: - Steel Grade A represents approximately 18% of raw material cost - Industrial Resin represents approximately 12% of raw material cost - Diesel represents approximately 8% of total cost base If any of these are in the flagged group, calculate the approximate margin impact of the price movement. If raw material costs are 40% of revenue, show what a 5%, 10%, and 15% rise in input costs does to gross margin percentage. Present as a simple sensitivity table within the document.
What to look for
  • Does Claude correctly apply the percentage assumptions to build the sensitivity table?
  • Is the table easy to read and act on?
  • Does Claude flag any limitations or assumptions it's making?
Step 3
Iteration — forward-looking briefing note
Follow-up prompt
Prompt to copy
Write a one-page market briefing note for the Head of Procurement / Commercial Director. Include: - Subject: Q3 2026 Input Cost Trend Review - Summary of the 13-week price movements (2-3 sentences) - A bullet-point "watch list" of the top 3 commodities to monitor in Q4, with a one-line rationale for each - Suggested hedging or procurement actions (e.g. forward contracts, spot buying windows, dual sourcing) - A note on Diesel as a transport cost input and what the trend means for logistics cost forecasting Keep it to one page. Tone: factual and commercial.
What to look for
  • Does the watch list prioritise correctly based on the data?
  • Are the procurement recommendations practical (not just generic advice)?
  • Is the note structured so a senior manager could act on it immediately?
💡
Extra challenge: Ask Claude to overlay two commodities on the same chart to compare their movement patterns. Then ask it to add a horizontal line showing the 13-week average for each. How does visualising the average alongside the trend make the anomalies easier to spot?
🔍
Independent Challenge A

The Anomaly Hunter

Multi-file analysis, anomaly detection, and presentation creation. Can Claude find the stories hidden in the data — and present them as slides?

Skills demonstrated

Multi-file analysis, anomaly detection, presentation creation (.pptx output)

All five files — in your cowork-demo-data folder
  • Q3-2026-Revenue-by-Division.csv
  • Q3-2026-KPIs.csv
  • Supplier-Variance-Report-Raw.csv
  • Q3-2026-Cost-Analysis.csv
  • Weekly-Market-Prices.csv
A

Find what the numbers are hiding

Don't tell Claude what to look for. Give it all five files and ask it to find the surprises. This tests whether Cowork can scan across multiple datasets, identify genuinely unusual data points, and explain why they matter — then package the findings as a slide deck you could present.

Prompt to try
Copy and send
Look at all the data files in my cowork-demo-data folder. Find the three most surprising or unusual data points across all five files. For each one, explain why it stands out, what the business implication might be, and what question you'd ask a manager about it. Present your findings as a short slide deck (3 slides, one per anomaly).
What a good result looks like
  • Claude produces a downloadable .pptx file — not just text. This tests Cowork's file creation capability.
  • The three anomalies are genuinely surprising (e.g. Asia Pacific revenue up 39.4%, a supplier 11.4% over contract, Direct-to-Customer Kits growing 32.9%) — not just the biggest numbers.
  • Each slide has a clear "so what" — not just "this number is high" but what it means for the business.
  • The manager questions are specific enough that you could actually ask them in a meeting.
💡
Why this matters: In real work, the most valuable thing AI can do with data is surface things you weren't looking for. This challenge tests whether you can use Cowork as an analyst who spots what you'd miss — and whether it can produce a presentation-ready output, not just text.
🎯
Independent Challenge B

The Board Question Prep

Data synthesis, executive communication, and question anticipation. Can Claude think like a CEO — and prepare you for the hard questions?

Skills demonstrated

Data synthesis, executive communication, question anticipation, Word document output (.docx)

All five files — in your cowork-demo-data folder
  • Q3-2026-Revenue-by-Division.csv
  • Q3-2026-KPIs.csv
  • Supplier-Variance-Report-Raw.csv
  • Q3-2026-Cost-Analysis.csv
  • Weekly-Market-Prices.csv
B

Anticipate the tough questions

The CEO will ask tough questions at the board meeting. Can Claude look at all the data, figure out what the hardest questions would be, and draft data-backed answers for each one? This is a different skill from analysis — it requires Claude to think adversarially about the data.

Prompt to try
Copy and send
The CEO will ask tough questions at the board meeting. Based on all the data in my cowork-demo-data folder, generate the 5 hardest questions a CEO would ask about Q3 performance, then draft a concise answer for each one backed by specific data points. Save as a Word document formatted as a Q&A brief.
What a good result looks like
  • Claude produces a downloadable .docx file — testing Cowork's Word document creation.
  • The questions are genuinely tough (e.g. "Why is on-time delivery below target while revenue is growing?", "What's driving the 7.7% overspend on warehousing and logistics?") — not softballs.
  • Each answer cites specific data points from the files, not generic statements.
  • The Q&A brief is formatted professionally enough to print and take into a meeting.
💡
Why this matters: One of the highest-value uses of AI in business is preparation — anticipating what you'll be asked and having the data ready. This challenge tests whether Cowork can shift from "analyse the data" to "think about the data the way a sceptical executive would."
📊
Independent Challenge C

The What-If Model

Financial modelling, scenario analysis, and Excel with working formulas. Can Claude build a spreadsheet you can actually use?

Skills demonstrated

Financial modelling, scenario analysis, Excel workbook creation (.xlsx with formulas)

Files needed
  • Q3-2026-Revenue-by-Division.csv
  • Q3-2026-Cost-Analysis.csv
C

Build a scenario model in Excel

This is the challenge that surprises most people. Cowork can create Excel workbooks with multiple tabs, working formulas, and structured financial models. Your challenge: get Claude to build a three-scenario model from the revenue and cost data — and produce a real .xlsx file you can open in Excel or Google Sheets.

Prompt to try
Copy and send
Using the revenue and cost data in my cowork-demo-data folder, build a scenario model in Excel. Create three tabs: - "Base Case" showing actual Q3 margins by division - "Bull Case" assuming 10% revenue growth and costs held flat - "Bear Case" assuming revenue flat and raw material costs up 8% Include a summary comparison across all three scenarios on a fourth tab. Use working formulas, not hardcoded numbers. I want to be able to change the assumptions and see the results update.
What a good result looks like
  • Claude produces a downloadable .xlsx workbook with 4 tabs — testing Cowork's spreadsheet creation.
  • The formulas actually work — open the file in Excel or Google Sheets and check that changing an assumption updates the outputs.
  • The summary tab compares margins across all three scenarios clearly.
  • The model is structured cleanly enough that a colleague could open it and understand it without explanation.
💡
Why this matters: Most people don't realise Cowork can produce working spreadsheets — not screenshots of tables, but actual .xlsx files with formulas. This is one of the most practically useful capabilities for anyone who works with financial data. If you only try one challenge, try this one.
💬
Independent Challenge D

The Natural Language Query

No structure. No instructions. Just a question in plain English. Can Claude figure out what you need from the data — and can it hold context when you follow up?

Skills demonstrated

Natural language querying, conversational context, data interpretation without explicit instructions

All five files — in your cowork-demo-data folder
  • Q3-2026-Revenue-by-Division.csv
  • Q3-2026-KPIs.csv
  • Supplier-Variance-Report-Raw.csv
  • Q3-2026-Cost-Analysis.csv
  • Weekly-Market-Prices.csv
D

Just ask a question

This is the simplest challenge — and the most revealing. Don't give Claude any formatting instructions, structural guidance, or output requirements. With all five files in your cowork-demo-data folder, just ask a plain English question. Then ask a follow-up based on the answer. This tests two things: can Claude interpret an ambiguous question using the data, and can it maintain context across turns?

Prompt to try
Copy and send — just this, nothing more
Which division is most profitable, and which supplier relationship should I be most worried about?
What a good result looks like
  • Claude identifies Acme Consumer as a strong growth story (Direct-to-Customer Kits up 32.9%, Consumer Packaged Goods up 17.2%).
  • Claude flags Northpoint Industries (10.0% over contract) or FlexForce Staffing (11.4% over) as the supplier relationships to worry about — and explains why.
  • The answer draws from multiple files without being asked to cross-reference.
  • Now send a follow-up question based on the answer (e.g. "What would you recommend we do about that supplier?") — does Claude maintain context, or does it lose the thread?
💡
Why this matters: The walkthroughs and earlier challenges taught you to write structured prompts. This challenge tests the opposite: what happens when you don't. In practice, the most common way people will use Cowork is by asking simple questions about their data. This challenge shows you how far that gets you — and where you might need to add more structure.
Final reflection — across all four challenges
  • You've seen Cowork produce different output types: slide decks (.pptx), Word documents (.docx), Excel workbooks (.xlsx), and conversational analysis
  • You've experienced the difference between structured prompts and open-ended questions
  • You've produced at least one output you could use in your real job with minimal editing
  • You've noticed at least one place where Claude's output needed correction — and you know how to prompt for it