Quarterly Report Pipeline

The Scenario

You're a regional sales manager preparing for the Q1 business review. Finance has sent you five CSV exports — revenue by product line, revenue by region, pipeline figures, customer acquisition costs, and headcount. Your VP's assistant has shared three PDF meeting notes from the quarter's leadership stand-ups. Your job: turn all of this into a polished report package that the VP can take into the board meeting on Friday.

Previously, this took you most of a day — wrangling spreadsheets, copying figures into slides, writing the narrative. Today, you're going to build a repeatable pipeline in Cowork that does it in under an hour.

What You'll Learn

By completing this tutorial, you'll be able to:

• Organise source data and meeting notes for optimal Cowork processing
• Write a detailed report specification before prompting — the single most impactful habit for report quality
• Craft an outcome-oriented delegation prompt that produces multi-format output (Excel + narrative)
• Spot-check quantitative outputs against source data to catch aggregation errors
• Evaluate qualitative synthesis (blending data insights with meeting note themes)
• Document a reusable pipeline that anyone on your team can follow

Prerequisites

• Claude Desktop with Cowork enabled (Pro or Max plan)
• Sample data files. You need CSV files and PDFs to work with. Options:
  • Use real (sanitised) quarterly data from your own organisation
  • Create 3-5 simple CSV files with fictional sales data (columns: Region, Product, Revenue, Target, Month) and 2-3 text files simulating meeting notes
• All files placed in a single working folder

Step 1: Organise Your Source Materials

Create a folder called Q1-Report-Pipeline on your desktop. Inside it, create two subfolders:

• source-data/ — place all your CSV files here
• meeting-notes/ — place your PDF or text meeting notes here

Leave the root of Q1-Report-Pipeline empty. This is where Cowork will write the output files. Keeping inputs and outputs separate is a professional habit that makes the pipeline repeatable — you always know where to find source materials and where outputs land, regardless of who runs the pipeline.

Your folder should look like this:

Q1-Report-Pipeline/
├── source-data/
│   ├── revenue-by-product.csv
│   ├── revenue-by-region.csv
│   ├── pipeline-figures.csv
│   ├── customer-acquisition-costs.csv
│   └── headcount.csv
├── meeting-notes/
│   ├── leadership-standup-jan.pdf
│   ├── leadership-standup-feb.pdf
│   └── leadership-standup-mar.pdf
└── (outputs will appear here)

Open each CSV briefly and confirm the data's clean — no broken headers, no empty rows at the top. Cowork handles messy data reasonably well, but garbage in still produces garbage out.

Things to check in your CSVs before you start:

• Headers in row 1 with no empty cells in the header row
• Consistent date formats (don't mix DD/MM/YYYY with MM/DD/YYYY across files)
• No merged cells or multi-line headers (common in Excel exports)
• Currency values should be numbers, not strings with currency symbols (£42,000 works; "forty-two thousand pounds" doesn't)
• No trailing empty rows or columns that could confuse aggregation

For your meeting notes PDFs, verify that the text is selectable (not scanned images). Cowork can read text-based PDFs directly but struggles with image-only scanned documents.

Checkpoint: Source materials are organised in subfolders. You've verified the CSVs are clean and the meeting notes are readable.

Step 2: Define the Report Specification

Before prompting Cowork, write down exactly what the finished report should contain. This is the specification that turns a vague "make me a report" into a delegatable task.

Your report package should include:

1. An Excel workbook (Q1-2026-Sales-Report.xlsx) with:

   • A summary sheet showing total revenue versus target by region
   • A product breakdown sheet with revenue per product line per month
   • A pivot-style analysis sheet cross-referencing region and product
   • At least one chart (revenue trend by month, or region comparison bar chart)

2. A summary document (Q1-Executive-Summary.md) with:

   • Three-paragraph executive overview
   • Top 3 wins from the quarter (drawn from meeting notes and data)
   • Top 3 concerns or risks
   • Key metrics table (total revenue, target attainment %, top product, top region)

Write this specification down before you prompt Cowork. You'll use it verbatim.

This step is where most people fail. They skip the specification and go straight to "make me a quarterly report." The result is something that looks like a report but doesn't match their expectations — wrong sheet structure, missing analysis dimensions, no charts, incorrect number formatting. Then they blame Cowork for "not understanding." It understood perfectly; it just wasn't told what to build.

The specification is your acceptance criteria. If the output matches the spec, the task succeeded. If not, you can point to exactly which requirement was missed and ask for a targeted correction rather than a full redo.

Checkpoint: You've got a written report specification covering both the Excel workbook and the summary document.

Step 3: Craft the Delegation Prompt

Translate your specification into an outcome-oriented Cowork prompt. The key principle: describe what "done" looks like, not how to get there.

Using the CSV files in source-data/ and the meeting notes in meeting-notes/, produce a Q1 2026 quarterly report package. Create two files in the root of this folder:

1. Q1-2026-Sales-Report.xlsx — An Excel workbook with four sheets: (a) Revenue Summary showing total revenue vs target by region, (b) Product Breakdown showing revenue per product line per month, (c) Regional-Product Matrix cross-referencing regions and products, (d) a chart sheet with a revenue trend line chart by month and a bar chart comparing regions.

2. Q1-Executive-Summary.md — A 2-page executive summary containing: a three-paragraph overview of the quarter, top 3 wins, top 3 concerns, and a key metrics table with total revenue, target attainment percentage, top-performing product, and top-performing region. Draw insights from both the data files and the meeting notes.

Use British English throughout. Format numbers with commas and currency as GBP (£).

Checkpoint: Your prompt includes the full specification with specific file names, sheet names, and formatting requirements.

Step 4: Submit and Review the Execution Plan

Point Cowork at your Q1-Report-Pipeline folder and submit the prompt. When Cowork presents its execution plan, review it against your specification:

• Does the plan mention reading all CSV files and all meeting notes?
• Does it list creation of both output files?
• Does it describe the chart types and sheet structure?
• Are there any steps that seem off — reading files outside your folder, creating files you didn't request?

If the plan looks incomplete (e.g., it mentions the Excel but not the summary document), don't approve it. Instead, ask Cowork to revise the plan to include all specified outputs.

Checkpoint: You've reviewed and approved an execution plan that covers all elements of your specification.

Step 5: Execute and Monitor

Click Allow and observe the execution. For this task, pay attention to:

• Data reading phase: Cowork should read all CSVs and extract the relevant figures. Watch for whether it processes them in parallel (each CSV is independent) or sequentially.
• Analysis phase: After reading, Cowork should cross-reference the data — calculating totals, comparing actuals to targets, identifying trends.
• Meeting notes integration: Cowork should read the PDFs and extract qualitative insights to blend with the quantitative data.
• Output creation: The Excel and Markdown files should be written last, after all analysis is complete.

This will likely take 3-8 minutes depending on file count and data complexity. Keep your machine awake and Claude Desktop open.

Watch for how Cowork handles the integration challenge — blending quantitative data from CSVs with qualitative insights from meeting notes. This is one of the harder tasks you can give it, because it requires switching between analytical reasoning (crunching numbers) and narrative synthesis (extracting themes from free-text documents). Many AI tools handle one or the other well; few handle both in a single workflow.

If you see Cowork process all the CSVs first and then the PDFs separately, that's a sign it may not integrate the two data types well. The best outcome is when the execution plan shows an explicit cross-referencing step where quantitative findings are enriched with qualitative context.

Checkpoint: Both output files have been created in the root of your working folder.

Step 6: Quality-Check the Excel Workbook

Open Q1-2026-Sales-Report.xlsx and audit each sheet:

• Revenue Summary: Are the totals correct? Cross-check one region's total against the source CSV manually. Spot-check at least two figures.
• Product Breakdown: Are all products and months represented? Are there any obvious gaps?
• Regional-Product Matrix: Does the cross-reference make sense? Do the row and column totals match the summary sheet?
• Charts: Are the chart types correct (trend line, bar chart)? Do the data points match the underlying sheets? Are axes labelled clearly?

Note any errors or quality issues. If something's wrong, tell Cowork what needs fixing — be specific about which sheet, which cell, what the value is, and what it should be. "The revenue total for London on the Summary sheet shows £420,000 but the source CSV totals to £438,000. Please recalculate." That's the right level of precision.

Common Excel quality issues to watch for:

• Rounding errors: Cowork may round figures differently than your standard convention. Specify in your prompt if you need "round to nearest pound" versus "two decimal places."
• Missing data rows: If one CSV had a different structure, some rows may have been dropped during the merge. Count the rows in the summary versus the sum of rows in the source files.
• Chart data ranges: A chart might reference the wrong cells, showing 11 months instead of 12 or missing a region. Click on the chart and verify its data source.
• Formula vs hardcoded values: Ideally, totals should use SUM formulas, not hardcoded numbers. This ensures the workbook remains usable if someone updates a figure. Check whether Cowork used formulas or just typed the calculated values.

Checkpoint: You've spot-checked at least three figures in the Excel workbook and verified the charts are correctly populated.

Step 7: Quality-Check the Executive Summary

Open Q1-Executive-Summary.md and evaluate:

• Accuracy: Do the key metrics match the Excel? The summary shouldn't contradict the data.
• Meeting notes integration: Did Cowork pull genuine insights from the meeting notes, or is the qualitative section generic filler?
• Tone: Is the language appropriate for a board-level audience? British English, no jargon, confident but not hyperbolic?
• Completeness: Are all specified sections present — overview, wins, concerns, metrics table?

If the summary feels disconnected from the data (a common issue when Cowork treats the CSVs and PDFs as separate tasks rather than an integrated analysis), prompt Cowork to revise specific sections with targeted corrections:

The executive summary mentions "strong performance in the Northern region" but the data shows Northern region missed its target by 12%. Please revise the summary to accurately reflect the regional performance data from the Excel workbook.

Specific, evidence-based correction prompts produce much better revisions than vague "make it better" requests. Always point to the specific inaccuracy and the source of truth.

Quality Scoring

Rate the executive summary on five dimensions:

A score of 4+ across all dimensions means the summary's ready to use. Below 3 on any dimension means that section needs another revision pass.

Checkpoint: The executive summary is accurate, integrates both data and meeting notes, and is board-ready.

Step 8: Document the Pipeline for Reuse

The real value of this tutorial isn't the report itself — it's the repeatable process. Create a file called pipeline-instructions.md in your working folder that documents:

1. Folder structure: Source data in source-data/, meeting notes in meeting-notes/, outputs in root
2. The exact prompt you used (refined after any corrections)
3. Quality checkpoints: The figures you spot-checked and the method you used
4. Known quirks: Any issues Cowork had with your data format, and how you worked around them

Next quarter, anyone on your team can drop new files into the same structure, paste the same prompt, and produce the same report package in minutes.

The pipeline document should be detailed enough that someone who's never used Cowork before could follow it. Include screenshots of the folder structure if helpful. Note the expected execution time so they know not to close their laptop. And emphasise the spot-check step — the pipeline only works if someone verifies the numbers before the VP sees them.

Checkpoint: You've got a documented, reusable pipeline that someone else on your team could follow without your help.

Expected Output

Your deliverable is a complete quarterly report package:

• Q1-2026-Sales-Report.xlsx with four analysis sheets and charts
• Q1-Executive-Summary.md with narrative, wins/concerns, and key metrics
• pipeline-instructions.md documenting the reusable process

This is the kind of output that takes a full working day when done manually. With a well-crafted pipeline, it takes under an hour — and produces consistent results every quarter.

The Compounding Value of a Pipeline

The first run through this pipeline takes the longest — you're calibrating prompts, discovering data quality issues, and refining the specification. The second run, next quarter, takes a fraction of the time. By the third quarter, the pipeline's mature: you drop in new CSVs, run the prompt, spot-check the numbers, and send it. The 8-hour task has become a 45-minute task.

More importantly, the pipeline produces consistent output. Every quarter's report follows the same structure, uses the same chart types, and applies the same formatting. When your VP opens Q2's report, it looks exactly like Q1's — she knows where to find the numbers, where the narrative lives, and what the charts show. Consistency breeds trust. A pipeline that produces unpredictable output each time forces your VP to re-learn the report format every quarter, which erodes confidence in the data itself.

Extension Challenges

1. Add a comparison dimension — Modify the pipeline to include Q4 data alongside Q1, producing year-over-year or quarter-over-quarter comparisons in the Excel and narrative.

2. Automate with scheduling — Set this up as a scheduled task that runs on the first Monday of each quarter. Place new CSVs in the source folder before the scheduled run, and Cowork produces the report automatically.

3. Add a presentation layer — Extend the pipeline to produce a 5-slide summary presentation alongside the Excel and narrative. Test whether Cowork can maintain consistent figures across all three output formats.