Most people use Excel the same way they always have: tab through cells, look up formulas, fix errors by hand. That approach is about to look very outdated.
Claude Cowork lets you describe what you want in plain English and get a working Excel file back. Clean this data. Write me a tiered commission formula. Turn this spreadsheet into a PDF report. No syntax. No Stack Overflow. No 45-minute tutorial just to figure out why your VLOOKUP broke.
This guide covers 10 real ways to use Claude Cowork for Excel work, from converting messy CSVs all the way through building charts, dashboards, and formatted reports you can actually send to a client.
What Is Claude Cowork for Excel?
Claude Cowork is the desktop app version of Claude that works directly with files on your computer. Unlike the web version, Cowork reads your local files, writes new ones, and saves them back to a folder you choose. It produces actual .xlsx files with working formulas, formatted headers, and multiple sheets. Not CSVs that need cleanup.
For Excel specifically, Cowork comes with a built-in spreadsheet skill that loads automatically whenever you ask for spreadsheet work. This skill knows Excel structure, formula syntax, data types, and common formatting patterns. You describe what you want and Cowork handles the technical execution.
The key difference from tools like Copilot in Excel: Cowork works on files you already have, in any format. You can hand it a text file, a Word doc, or a messy CSV and it converts it into a clean structured spreadsheet. No add-in required. No account linking.
See also: Claude Cowork setup guide
See also: Claude Cowork Projects
1. Convert Any File Format Into a Clean Excel Spreadsheet
Claude Cowork can take a file in any format and turn it into a structured .xlsx file. PDF, CSV, text file, Word document. Drop the file into your Cowork folder and give it a plain English prompt. The output is a formatted spreadsheet with proper column headers, consistent data types, and no manual cleanup.
Converting a Messy CSV
Start with a CSV that has broken formatting, missing headers, or inconsistent column order. In Cowork, type a prompt like this:
- ‘Take the invoice sample CSV and turn it into a clean formatted Excel spreadsheet with proper headers and formatting.’
Cowork reads the file, identifies the column structure, applies consistent formatting, adds a totals row where appropriate, and saves the new .xlsx to your folder. The output includes clean headers, formatted currency columns, and a status field with values like paid, pending, or adjustment. Work that would take 15-20 minutes manually takes under two minutes.
Converting a Text File to Spreadsheet
Unstructured text files are harder because there are no column delimiters to rely on. A contact list might have names and emails in a paragraph format, separated by commas, pipes, or inconsistent spacing.
A prompt like ‘The team contact text file has a list of contacts. Convert it into an Excel spreadsheet with columns for name, email, phone, and department.’ handles this. Cowork reads the file, identifies up to five different delimiter types in a single file, handles rows where columns appear out of order, and produces a clean spreadsheet. A 10-contact file with five different delimiters and swapped column positions comes out clean with no manual intervention.
2. Understand an Inherited Spreadsheet in Minutes
Inheriting a spreadsheet from a coworker is one of the most frustrating experiences in Excel. Formulas referencing other sheets, columns with no labels, logic that only made sense to the original author. Claude Cowork can read the whole file and explain it in plain English before you even open it yourself.
Get a Plain English Explanation
Drop the spreadsheet in your Cowork folder and ask: ‘I just inherited this spreadsheet from a coworker. Can you explain what it does in plain English?’
Cowork returns a structured breakdown: which sheet is the main data sheet, what each column tracks, how the key formulas work, and what values drive the calculations. For a sales tracker, that means knowing the commission rate column, what the ‘above target’ flag means, and which values feed the summary sheet.
Walk Through Every Formula
Follow up with: ‘Walk me through the formulas in this file. What is each one calculating and why?’
Cowork maps each formula to its column and explains the logic in plain terms. Column G totals quarterly sales across four columns. Column I multiplies revenue by commission rate. For a rep at $60K revenue and a 10% rate, Cowork shows you the math: $6,000 earned. It also flags anything worth flagging, like a hardcoded value in a summary cell that won’t update automatically if the underlying data changes. That kind of error is easy to miss when you’re reading raw cell references.
3. Audit and Fix Data Errors Automatically
Spreadsheet errors aren’t always obvious. A date formatted as 1900 (Excel’s zero date, meaning an empty cell was misread as a number). A duplicate order ID. A negative quantity. A customer name that’s actually a string of numbers. These hide in plain sight across large files and take hours to catch manually.
Detecting Errors with One Prompt
The prompt ‘Review the spreadsheet and flag anything that looks like an error, anomaly, or red flag’ is enough. Cowork scans the entire file and returns a numbered list of every issue it finds.
For a 20-row order sheet, Cowork found six issues: a duplicate order ID, a future date three years ahead of the data range (almost certainly a typo), a corrupted date in the 1900 format, a negative quantity, a customer name that was a string of numbers, and a row where the total didn’t match the quantity times the unit price. That last one is the kind of error that’s easy to miss when auditing manually, because the math looks right until you check the formula.
Auto-Fix and Save a Clean Version
After flagging errors, Cowork asks: ‘Want me to fix all of these and save a cleaned version of the file?’ Reply with ‘Yes, fix the file and save a second version called clean_[filename].’ Always save to a new file name. Never overwrite the original when fixing data issues.
Cowork applies all six fixes, saves the clean version, and provides a change log. The duplicate ID gets renamed (e.g., order 2 becomes order 2B). The future date gets corrected. The corrupted date gets cleared with a ‘review required’ note. Negative quantity is corrected. Customer name flagged for manual review. Math error recalculated. Ask for a change log on a second tab when fixing larger files. It makes the audit trail easy to follow.
4. Clean Messy Data in Under Two Minutes
Inconsistent formats are one of the most tedious parts of any data project. State names written as NY, New York, and New York (with a trailing space). Phone numbers in five different formats. Emails in all caps. Dates formatted three different ways in the same column. Claude Cowork standardizes all of it from a single prompt.
The Cleaning Prompt
‘Clean up this spreadsheet. Standardize all the formatting, fix capitalization, and make it consistent throughout.’
Cowork scans the file and returns a plan before touching anything: what it found (capitalization issues, inconsistent phone formats, mixed date formats, blank industry values, three duplicate rows) and what it intends to do. It then asks two clarifying questions before proceeding.
Clarifying Questions Cowork Asks
For any values it can’t infer, Cowork asks rather than guesses. For blank industry values: ‘How should I handle eight rows with a missing industry value? Fill as Unknown or leave blank?’ For phone numbers: ‘What phone format should I standardize to?’ Answer those and Cowork runs the full cleanup.
The output on a 30-row contact list: zero nulls, zero duplicates, 27 clean rows (three duplicates removed). Names fixed to title case. Emails lowercased. All phone numbers in (XXX) XXX-XXXX format. All state abbreviations to two characters. All dates in a consistent MM/DD/YYYY format. Colored header row applied. A 30-minute manual cleanup job done in under two minutes, and Cowork found the duplicate rows without being asked.
See also: Claude Cowork tutorial
See also: Claude Cowork tips and tricks
See also: Claude Cowork PowerPoint guide
See also: Claude Cowork web scraping guide
Need Help Building AI Automations?
Join Our AI Community
5. Write Excel Formulas in Plain English
Formula syntax is one of the biggest barriers to using Excel effectively. Most people know they need a formula but spend 20 minutes on Stack Overflow figuring out the exact syntax. Claude Cowork removes that step entirely. Describe what you want in plain English and Cowork writes the formula, applies it to the correct column, and handles edge cases like error trapping.
Five Formulas Cowork Wrote Without Syntax Help
Starting with a spreadsheet that has columns for sale ID, rep name, date, product, category, revenue, cost, and discount, here are five prompts that produced working formulas:
- ‘Add a profit margin column that calculates the margin as a percentage of revenue.’ Cowork adds the formula =(F2-G2)/F2 with IF(ISERROR()) wrapping for error handling. Most people would skip the error handler.
- ‘Create a running total column that shows the cumulative revenue as you go down the rows.’ Cowork uses =SUM($F$2:F2) and adjusts the reference as it copies down.
- ‘Add a column that flags any sale where the discount is greater than 20% with the word review.’ Produces an IF formula checking the discount column with a text output.
- ‘Write a formula that calculates how many days ago each sale was made.’ Uses =TODAY()-DATEVALUE() with the correct cell reference.
- ‘Add a commission column: 8% for sales under 5K, 10% for 5K to 10K, and 12% for anything else.’ Produces a nested IFS formula covering all three tiers. Getting the bracket logic right manually takes a few tries for most people.
All 25 formulas across five columns verified with zero errors. The error handling on the profit margin formula alone (which most people skip) is the kind of best-practice detail that separates clean work from work that breaks when data is messy.
6. Build and Manage Multi-Sheet Workbooks
When your data lives across multiple sheets, pulling it together into a summary view usually means writing SUMIF or INDIRECT formulas and making sure every range reference updates when new rows are added. Claude Cowork handles this from a single prompt.
Creating a Cross-Sheet Summary
With four regional sheets (north, south, east, west) each containing monthly revenue data, the prompt ‘I have data on separate sheets for each region. Create a new summary sheet that pulls the total revenue from each region into one table’ produces a summary sheet with a region column and monthly revenue breakdown.
Making the Summary Update Automatically
Hardcoded range references break as soon as new rows are added. Follow up with: ‘If I add a new row to any regional sheet, will the summary update automatically? If not, fix it.’
Cowork checks every formula in the summary sheet. If any ranges are hardcoded (e.g., =SUM(North!B2:B20) instead of a dynamic reference), it replaces them with formulas that expand as the data grows. The fix takes one prompt and about 30 seconds.
7. Analyze Data Without Pivot Tables
Pivot tables are powerful but have a learning curve. For most analysis questions, you don’t need them. Claude Cowork can read a dataset and answer specific business questions directly in the chat, then output formatted summaries you can paste into an email or Slack message.
Plain English Data Questions [H3]
Starting with an e-commerce dataset of 50 orders across six months with columns for product, category, customer tier, day of week, and revenue, here are the analysis prompts and what they returned:
- ‘Summarize the key trends in this data set. What stands out to you?’ Cowork identified: electronics accounts for 60% of total revenue, June had a significant revenue spike, gold tier customers drive the most order volume, keyboards are the top product, and order values have high variance (mean $305, standard deviation $376), meaning a small number of high-value orders are skewing the average.
- ‘Who are the top five customers by total spend?’ Cowork returns a ranked table with total spend, order count, average order value, and tier. It also flags an anomaly: three of the top five spenders are bronze tier, which raises a question about how tier assignment works.
- ‘Is there a pattern in which days of the week have the highest sales?’ Friday and Saturday are peak days. Wednesday is the lowest. Sunday has the most orders but underperforms on revenue.
Executive Summary in One Prompt
‘Write me a three-sentence executive summary on this data I could paste into a Slack message to my manager.’
Output: ‘Across 50 orders from January through June, electronics and gold tier customers drove the majority of revenue, with Friday and Saturday alone accounting for 58% of total sales. The top customer contributed $3,650 individually. Three of the five highest spenders are bronze tier, suggesting the current tier classification may not be aligned with actual spend behavior.’
One prompt. Ready to paste. No pivot table.
8. Create Charts and Dashboards Directly in Excel
Charts and conditional formatting are where a lot of people’s Excel skills hit a ceiling. Cowork handles both from plain English prompts, producing actual embedded charts and formatting rules inside the .xlsx file.
Charts from Plain English
With a performance dataset containing monthly revenue, expenses, profit, leads, and conversions:
- ‘Create a bar chart showing total revenue by month.’ Cowork generates an embedded bar chart in the spreadsheet.
- ‘Add a line chart that shows both revenue and expenses on the same chart so I can see the trend.’ Cowork produces a dual-series line chart with revenue as a solid line, expenses as a dashed line, and the profit gap shaded between them.
Conditional Formatting and Dashboards
‘Apply conditional formatting to the profit column and add a color scale to the conversion rate column.’ Both formatting rules are applied in a single response.
‘Create a simple dashboard view on a new sheet that shows key metrics at a glance.’ Cowork adds a new sheet with a header banner, KPI summary row, and the full data table with color-coded columns. If you want an HTML dashboard instead of an in-spreadsheet one, use the prompt ‘Create an HTML dashboard from this data.’ The HTML version is easier to embed in reports or internal tools.
9. Format Any Spreadsheet for Professional Presentation
A spreadsheet that works isn’t always a spreadsheet that looks good. Raw data dumps have no formatting, inconsistent column widths, and numbers without currency symbols. When you’re sending a file to a client or a manager, it needs to look like it was designed, not dumped from a database.
The prompt ‘Make this spreadsheet look professional. Clean up the formatting, add proper headers, and make it presentable.’ takes an ugly but accurate file and returns a formatted version.
For a project budget spreadsheet with columns for department, allocated budget, actual spend, variance, and status, Cowork applies a clean header row with color, consistent column widths, currency formatting on numeric columns, and a status column with visual indicators. The columns are renamed to be cleaner (‘Allocated’ instead of ‘budget_alloc’). The output is something you could send to a client without apologizing for how it looks.
From there, refine with targeted prompts: change the column headers, adjust the title, update the color scheme. Each iteration takes one prompt and the file updates in place.
10. Export Excel Data to Reports and Other Formats
Claude Cowork doesn’t just work inside Excel files. Once your spreadsheet is clean and analyzed, you can transform it into other formats without leaving the desktop app. PDF reports, Word documents, written summaries, email drafts. The file is the input; any format is the output.
Converting a Spreadsheet to a PDF Report
The prompt ‘Turn the spreadsheet into a formatted PDF report with a title, executive summary, and the data table.’ is enough to get a structured PDF.
For a department headcount and budget sheet, Cowork produces a PDF with a title banner (‘Q4 Department Performance Report, Fiscal Year 2025’), a KPI summary row, a formatted data table, and a confidentiality footer. The report is production-ready out of the box.
If the layout needs adjusting (e.g., a table is split across pages), a single follow-up prompt handles it: ‘I want the full data table on the second page. Don’t split it across pages.’ Colors, fonts, and section order are all adjustable the same way. One prompt per change.
See also: Claude Cowork skills
Three More High-Value Use Cases
The 10 sections above cover what’s in the video. But there are three use cases the Claude docs specifically highlight that are worth knowing about, especially if you work in finance or operations.
Budget vs. Actuals Variance Analysis
One of the most common recurring tasks in any business is comparing what you planned to spend against what you actually spent. The manual version means pulling two sheets together, writing SUMIF formulas, and formatting a variance table every single month.
With Cowork, the prompt is: ‘I have a budget sheet and an actuals sheet for Q1. Build a variance table that shows each department’s allocated budget, actual spend, the dollar variance, and the percentage variance. Highlight anything over 10% in red.’
Cowork builds the cross-sheet variance table, applies the conditional formatting, and calculates both dollar and percentage variance in one pass. The output is ready to paste into a board deck or email to your manager. For teams that run this monthly, saving the prompt and dropping in a new actuals file each month is all it takes.
Month-End Close and Bank Reconciliation
Drop your bank statement and your internal transaction log into your Cowork folder and ask it to reconcile them. Cowork reads both files, matches transactions by amount and date, flags discrepancies, categorizes expenses by vendor, and produces a formatted reconciliation report.
The prompt: ‘Reconcile the bank statement against the internal transaction log. Flag any discrepancies, categorize expenses by vendor, and produce a formatted reconciliation report with a summary of unmatched items.’
One thing the Anthropic docs note: the more you run this, the better the output gets. On the first run, Cowork flags everything uncertain. By the second and third month, it applies the patterns it identified in prior runs and handles the straightforward matches automatically. The exception list gets shorter each time. Finance teams running monthly closes have reported cutting the reconciliation time from a half day to under an hour.
Always save the original files before running. Ask Cowork to produce a change log on a separate tab so you have a full audit trail of what was matched and what was flagged.
Scenario Analysis and Sensitivity Tables
Scenario analysis is one of the more time-consuming parts of financial modeling. Building a base, bull, and bear case means duplicating sheets, adjusting assumptions, and making sure every formula updates correctly across all three versions.
Cowork handles this from a single prompt: ‘Build a scenario analysis with base, bull, and bear cases using the revenue assumptions in column C. Base is current values. Bull is 20% higher. Bear is 15% lower. Show the impact on net profit for each scenario in a summary table.’
Cowork creates the three-scenario model, links the assumptions to the revenue and profit formulas, and adds a summary table that shows side-by-side outputs. Changing a single assumption updates all three cases simultaneously. For sensitivity tables (e.g., how does margin change as both price and volume shift?), use: ‘Build a sensitivity table showing net margin across five price points and five volume levels.’
Tips and Limitations Worth Knowing
Use Change Tracking for Audit Trails
Any time Cowork edits a file you’ll share with others, ask it to track changes with comments. The prompt: ‘Apply all the changes and add a comment to each modified cell explaining what you changed and why.’
This creates an inline audit trail inside the Excel file. Every edited cell shows who changed it (Claude), when, and the reason. For files going to a client or a manager, this turns a black-box edit into a transparent one.
Ask Questions About the Workbook — It Cites Specific Cells
Cowork doesn’t just edit spreadsheets. You can ask it analysis questions and it returns answers with clickable citations pointing to the exact cells it referenced. ‘Why does the profit margin drop in March?’ will return an explanation that references the specific cells in March’s revenue and cost columns, not just a generic answer.
This is useful for auditing models you didn’t build. Instead of tracing formulas by hand, ask Cowork to explain the logic and it maps every claim back to its source cell.
Set Persistent Instructions Once
If you have formatting preferences that apply across every spreadsheet you work on, set them as persistent instructions in your Cowork settings rather than repeating them in every prompt. Examples: ‘Always use title case for column headers.’ ‘Always add a totals row to any data table.’ ‘Always save a backup copy before making changes.’
Once set, Cowork applies these rules automatically without being asked. It’s the equivalent of a house style guide for your spreadsheets.
What Cowork Cannot Do in Excel
Three things worth knowing before you hit a wall mid-workflow:
- VBA and macros: Cowork can open .xlsm files but cannot read or write macro code. If your workflow depends on VBA automation, Cowork won’t touch that layer.
- Power Query: Not supported. If your spreadsheet uses Power Query to pull data from an external source, Cowork can work with the output data but can’t modify the query itself.
- Live data connections: Cowork works with static file snapshots. It doesn’t connect to live databases or refresh external data links.
Only Use Cowork With Spreadsheets You Trust
Anthropic’s own documentation flags this: spreadsheets from untrusted external sources can contain hidden instructions embedded in cells, formulas, or comments designed to manipulate Claude’s output. This is called prompt injection.
The practical rule: if a file came from a stranger on the internet, don’t drop it straight into Cowork without reviewing the cells first. Files from your own systems, coworkers, or known vendors are fine.
Frequently Asked Questions About Claude Cowork and Excel
Can Claude Cowork edit an existing Excel file or only create new ones?
Claude Cowork can both create new Excel files and edit existing ones. Drop any .xlsx or .csv into your selected Cowork folder and ask it to make changes. Cowork reads the current file, applies the requested edits, and saves the updated version. For risky operations like data cleaning or error fixes, always ask it to save a new version with a different filename rather than overwriting the original.
Does Claude Cowork work with Excel files that have multiple sheets?
Yes. Cowork reads multi-sheet workbooks and can create cross-sheet formulas, build summary views, and fix hardcoded range references that break when new rows are added. It understands the relationship between sheets and can update summary formulas to reference all sheets dynamically.
How accurate are the Excel formulas Claude Cowork writes?
Cowork verifies formulas after writing them and reports an error count when it’s done. In most cases you’ll see ‘zero errors across all formulas’ in the output. For complex nested formulas like tiered commission structures, Cowork also adds error handling (IFERROR wrapping) that most users skip when writing formulas manually. Always spot-check at least one row of calculated output to confirm the logic is correct.
What file formats can Claude Cowork convert into Excel?
Cowork can convert CSV files, plain text files, Word documents, and PDFs into Excel spreadsheets. It handles inconsistent delimiters, columns in the wrong order, and mixed data types. For text files, it infers the column structure from context and asks clarifying questions when it’s not obvious.
Does Claude Cowork work with Google Sheets?
Claude Cowork produces .xlsx files, not Google Sheets files. However, any .xlsx file Cowork creates can be uploaded directly to Google Sheets. If you want to verify the output in Sheets, download the file from your Cowork folder and import it. All formulas, formatting, and multiple sheets transfer correctly on import.
Start Using Claude Cowork for Your Spreadsheet Work
The 10 approaches in this guide cover the most common Excel tasks that eat time every week. Converting formats, cleaning data, understanding inherited files, writing formulas, building summaries, analyzing without pivot tables, creating charts, formatting for presentation, and exporting to reports.
The pattern is always the same: choose your folder, drop in a file, describe what you want in plain English. Claude Cowork handles the technical execution. You review the output and refine with follow-up prompts if needed.
Start with one task you do repeatedly. If you spend 30 minutes every Monday cleaning a data export, that’s the first thing to try. Open Cowork, drop the file in, and type what you want. The time savings stack up fast.
- Download the Claude desktop app and open the Cowork tab
- Choose a folder on your computer where your data lives
- Drop in any spreadsheet, CSV, or text file
- Describe what you want in plain English and click send
- Review the output, refine with follow-up prompts as needed
