9 steps in Claude Cowork to automate your investment analysis

In 15 minutes to a full investment memo

I just got Claude to do an entire equity research workflow on Adyen in under 15 minutes: pull the data, build the Excel model, run a 3-scenario DCF, write the investment memo, and even schedule a weekly update task. No code. Just prompts. Below you’ll find the exact 9-step playbook.

I have been using Claude Code heavily to develop investing tools to track my investments and research a slew of companies.

But I can understand that most of you don’t want to open a terminal, start coding, and spend a couple of hours without a guaranteed result.

Claude Cowork is sort of Claude Code, but for non-developers.

But the distinction is the following:

• Claude Code is for building and automating. It’s the most versatile of the bunch but comes with a steeper learning curve.

• Claude Cowork also lets you automate tasks, sort of

That’s what we’re diving into today.

How to use Claude Cowork to analyze stocks.

As Claude himself so neatly puts it:

• Chat is for thinking

• Code is for building

• Claude Cowork is for doing

What Claude Cowork can do that Claude projects can’t.

We talked about Claude’s projects in a previous article, but Cowork goes beyond its capabilities.

Here are some examples of what Cowork can do that Projects cannot:

• Read and write files in a folder

• Build Excel files with formulas

• Scheduled tasks that run things on auto

• Use plugins: Like financial or equity research

So what would this look like in practice? Let’s build a stock analysis and first valuation model within Cowork. We’ll take the company Adyen as an example. (for our full process, you can check it here)

Step 1: Set up the scaffolding

We’re going to give Claude Cowork access to a folder on our computer. It can create, delete, and modify files.

So we created a stock research folder and added it to the chat.

The next step is to create a structure for each company that Cowork will use. Here’s an example of a structure for Adyen that will live on your computer.

---

ADYEY/

├── filings/ (annual reports + press releases)

├── transcripts/ (earnings calls)

├── charts/ (weekly chart)

├── model/

│ ├── adyen_financials.csv

│ └── ADYEY_model.xlsx ← real Excel model

└── memo/

├── 01_business_overview.md

├── 02_technical_read.md

├── 03_thesis.docx ← investment thesis

└── 04_one_pager.pdf

---

It’s easy.

Just ask Cowork to set up a folder structure, then copy and paste the structure above.

In a matter of seconds, Cowork will build this:

Step 2: Download the data

We now have to fill out the folders. That means downloading annual reports and earnings transcripts and getting our hands on financial data.

For financials, you have 2 options:

• Free: Use finance (see prompt below)

• Paid: Connect your data provider of choice

Go to Customize, Connectors, and you can find everything finance-related by filtering

For the purposes of this article, I’m going to assume you don’t have a data connector, so we’ll use yfinance to illustrate.

Now just paste the following prompt and let Cowork do its magic:

---

You have access to my ADYEY research folder. Please gather the following source material for Adyen (ticker ADYEY, primary listing ADYEN.AS on Euronext Amsterdam) and save everything into the correct subfolders. Work through the steps in order and tell me if anything fails so I can help.

1. Annual reports (last 5 years) Go to https://www.adyen.com/investor-relations and download Adyen’s five most recent annual reports as PDFs. Save them in filings/ using the naming format ADYEN_Annual_Report_YYYY.pdf.

2. Quarterly / half-year reports (last 4) From the same investor relations page, download Adyen’s four most recent shareholder letters or half-year / trading update reports. Save them in filings/ as ADYEN_Report_YYYY_HX.pdf or ADYEN_Trading_Update_YYYY_HX.pdf depending on the document type. Note: Adyen reports semi-annually, not quarterly, so you’ll mostly find H1 and FY reports plus trading updates — grab the four most recent of whatever is available.

3. Earnings call transcripts (last 4) Find the four most recent Adyen earnings call transcripts. Try Adyen’s investor relations page first, then fall back to public sources like Seeking Alpha, Motley Fool, or Roic.ai. Save each one as plain text in transcripts/ using the format ADYEN_Transcript_YYYY_HX.txt. If a transcript isn’t publicly available, note which one is missing in a file called transcripts/README.md.

4. Financial data via yfinance (last 5 years) In the Python sandbox, install yfinance if needed and pull the following for ticker ADYEY:

• 5 years of daily price history (Open, High, Low, Close, Adjusted Close, Volume)

• Annual income statement

• Annual balance sheet

• Annual cash flow statement

• Quarterly income statement, balance sheet, and cash flow statement

• Key info dictionary (market cap, shares outstanding, currency, sector, etc.)

Save each dataset as a separate CSV in model/:

• ADYEY_prices_5y.csv

• ADYEY_income_annual.csv

• ADYEY_balance_annual.csv

• ADYEY_cashflow_annual.csv

• ADYEY_income_quarterly.csv

• ADYEY_balance_quarterly.csv

• ADYEY_cashflow_quarterly.csv

• ADYEY_info.csv

5. Summary When you’re finished, give me a short summary listing exactly which files you saved, which (if any) you couldn’t find, and flag any data quality issues you noticed in the yfinance pull (missing years, currency mismatches, etc.). Do not start analysis yet — this step is just data collection.

---

You can see Cowork doing this exactly as asked. It takes about 5 minutes to fetch all the data.

Here’s a glimpse of what’s inside the annual reports folder:

Note: It took 5 minutes, mainly because it was trying to fetch the earnings transcripts. If you want these faster and more reliably, you’ll need a paid provider, such as Financial Modeling Prep, which has an MCP server. Cowork will then be able to use that one.

So we did the setup, we have the data, let’s continue.

Step 3: Let Claude understand the business

Now that we have everything set, Claude needs to ingest the actual information.

Here’s the prompt:

---

“Read all PDFs in filings/. Write a one-page business overview of Adyen in memo/01_business_overview.md covering: what they do, how they make money (take rate, processed volume, net revenue), main segments and geographies, top customers and customer concentration, competitive position vs. Stripe / PayPal / Block / Worldline, and the three biggest structural risks. Use plain English, no jargon.”

---

It generated this business overview markdown file, which I think provides a good basic coverage of what Adyen does.

Step 4: Turn the yfinance CSV file into an Excel

The data from yfinance just sits in several CSV files. We want to build a solid Excel out of it.

Here’s the prompt:

---

“Using the CSVs in model/, create model/ADYEY_historicals.xlsx with three tabs: Income Statement, Balance Sheet, Cash Flow. Show the last 5 years side by side. Below each statement, add a ‘Key Ratios’ block calculating: revenue growth YoY, gross margin, EBITDA margin, FCF margin, net cash position, and ROIC. Format numbers in € millions, one decimal. Flag any line item where yfinance data looks incomplete or inconsistent.”

---

So Cowork created an Excel, and a summary of what it did, and you immediately see the power of Opus 4.6

So, per point 3, we ask Claude to update the Excel file and include the float.

It updated the existing Excel with this:

caption...

Despite the minor naming error, this is important if we want to work with the correct enterprise value and free cash flow later on.

Claude did warn us to cross-check the data against the annual reports, as the actual cash calculator is an approximation based on payables.

Of course, we’re not going to do that. We just asked Claude to do it for us.

So its approximation was only slightly off. It then proceeded to update the Excel.

Step 5: Build a valuation model

We’ll build a simple 5-year DCF and comparables tab with the following prompt:

---

“Add two new tabs to model/ADYEY_historicals.xlsx: ‘DCF’ and ‘Comps’.

DCF tab: 5-year forecast with three scenarios in separate columns — Bear (10% revenue CAGR, EBITDA margin flat at current level), Base (18% CAGR, margin expanding to 55% by year 5), Bull (25% CAGR, margin to 60%). Assume 22% tax rate, 9% WACC, 3% terminal growth. Calculate enterprise value, equity value, fair value per share in EUR, fair value per ADYEY ADR in USD (use a sensible EUR/USD rate and note it), and upside vs. today’s price.

Comps tab: Pull in Stripe (note: private, use latest funding round valuation), Block (SQ), PayPal (PYPL), Nuvei (NVEI), and Worldline (WLN.PA). For each, show market cap, EV, revenue, EBITDA, EV/Revenue, EV/EBITDA, and revenue growth. Apply Adyen-appropriate multiples to get a comps-based fair value.

All formulas should be live, not hardcoded. Add a ‘Summary’ row at the top of each tab showing the fair value conclusion.”

---

Now, if you look at this step, you could also fetch the data from the comparables in the previous data step. That’s a choice.

So the DCF Excel includes actual formulas. So you can adjust the WACC or the terminal growth rate to better align with your own assumptions.

Based on this analysis, the overall case is bullish towards Adyen.

Let’s look at the comparables table:

Stripe is the most comparable company. Of course, Stripe is valued higher, and we need to be careful as it’s not a public company.

Step 6: Create the investment memo

Now, let’s create the final investment memo from all of the above.

Use the following prompt:

---

“Read memo/01_business_overview.md, memo/02_key_questions.md, and the DCF + Comps tabs in model/ADYEY_historicals.xlsx. Write a 2-page investment thesis in memo/03_thesis.docx with these sections:

1. Thesis in one paragraph

2. Why now (catalysts)

3. Key drivers of the bull case

4. Valuation summary (base case fair value, upside, comps check)

5. Key risks

6. What would make me wrong (the anti-thesis)

7. Position sizing and suggested entry range

Tone: professional but conversational, like a buy-side analyst writing to a PM. No hype, no fluff.”

---

Now, of course, you can use your own framework for an investment memo. Even better would be to provide your own MD file with your typical memo framework.

You can download the thesis memo below.

Investment Memo Adyen

417KB ∙ PDF file

Download

Download

Conclusion: Adyen is a buy right now.

Step 7: Bonus: Create a one-page summary

The memo is not that long, but if you want a single page with only the signal, use this prompt:

---

“Create memo/04_one_pager.pdf: a single page with ticker, current price, base-case fair value, upside %, 3 bullets bull case, 3 bullets bear case, and the valuation summary table from the DCF tab. Clean, skimmable design.”

---

You’ll get something like this:

Step 8: Track the investment

So, we have a single photo for Adyen today. But what if new earnings come out?

You just set up a scheduled task within Cowork.

---

“Create a scheduled task called ‘ADYEY Update’ that runs every Monday at 7am. On each run it should: (1) pull fresh price data via yfinance and update the historicals file, (2) check adyen.com/investor-relations for any new filings since the last run and download them, (3) if a new earnings release is found, update the model with the new quarter and write a short memo/earnings_update_YYYY_MM_DD.docx comparing results to my thesis in memo/03_thesis.docx, and (4) email me a one-paragraph summary of what changed.”

---

Now, bear in mind that your computer needs to be on and Claude needs to be activated. This is not a task that will work all the time. You could, for example, adjust the timing so it runs when you’re working on your laptop (maybe at 9 AM on a Monday).

(If you want tasks that run all the time, you need an always-on machine like a Mac mini, where your Claude Cowork is open. Or you can do it like me, and have a Mac mini where a script built with Claude Code does all of this every week.

Conclusion

These 8 steps provide you with an investment memo and a scheduled task for a single company.

You’ll need to decide for yourself if you’re happy with it. And iterate if needed.

What’s next?

Now comes the real power of Cowork and Claude. We did this once for Adyen. We don’t have to repeat all the steps for another company.

You can build a skill from this and run it at any company you want. Cowork will perform all the steps in this article one by one. You just need to point it to a directory where it can store and process the files.

Bonus step 9: Building the skill using the skill-creator skill

Yeah, it’s a bit meta. But it works. Claude has a skill to create skills. (We covered this before)

Just use this prompt:

---

“Use the skill-creator skill to turn my Adyen research workflow into a reusable skill called equity-research. The skill should take a ticker symbol as input and reproduce everything I just did for ADYEY: create the folder structure, pull yfinance data, build the historicals Excel, draft the business overview, key questions, DCF and comps model, thesis, and one-pager. Use the files in ADYEY/ as the reference example of what the output should look like. Keep the DCF assumptions parameterizable so I can adjust them per company.”

---

And it will create an equity research skills file in your folder.

Now go to Customize > Skills and add the file to your skill list. Once you do, you’ll see this:

Now, as a test, I ran the skill on another company, for example, IREN.

But before blindly executing the skill, always ask Claude the plan first:

Use the equity research skill to analyze a company called IREN. Show me the plan first before executing

The advantage is that before it actually starts using the skill, you can tweak it a bit.

Now, if you know nothing about this company, it will be difficult.

In this case, let’s just run it as a default skill run and see what it does.

Once Claude is running the skill, you’ll see how it advances through all the steps on the right side of your screen:

Note: I made a small modification to this skill to fetch annual reports from EDGAR, since this company is US-listed.

SO it took a couple of minutes, and now I have this:

With a memo and valuation model for IREN.

It considers IREN to be fairly valued, with a wide variation in possible future outcomes. That makes sense.

Final Conclusion

The bottom line is you can build things fast. And compound on the things you build.

The real work is the iteration and refinement to achieve the quality you need.

But once you’re past that hurdle, you can automate everything.

But always be mindful.

Claude is excellent at using your frameworks and your assumptions to give you a clear picture of a company.

But do not let it judge. You’re the judge.

Keep learning AI. Keep building.

Kevin