Hey Reader,

3 days → 2 hours.​
One of our AI Finance Club member used to burn three days every week tagging cryptic credit-card charges like “AMZN*3F8K.”

But then she trained a Custom GPT on her chart of accounts and vendor quirks; now the same job takes two hours and she’s back to high‑value work.

Sound familiar? It could be you—or the unlucky soul on your team.

This Wednesday, 13 August, I’m going to show you similar use cases (register here).
I’ll demonstrate exactly how finance teams are automating close processes, step by step, with no coding required. Just one tool, a few smart prompts, and the power of AI.

Your Free AI Masterclass

Mastering AI in Accounting

​👉Reserve your free spot now👈 ​

In 1 hands-on hour you’ll learn:

• Build your CustomGPT to automate tasks
• Clean Data with AI
• Financial Analysis
• and more practical use cases!

👉Save your free seat​

This masterclass is sponsored by Zenskar.

Expense Categorization with Custom GPTs

​

Expense Categorization demo starts at 4:31.

Below is the step-by-step method for building your own Custom GPT that categorizes expenses, just like we discussed before in the beginning of this email.

These are the practical instructions you can apply immediately.

If you want to learn it live, don't forget to come to my next masterclass on Wednesday, 13 August.

Enjoy!

Step 1: GPT Creation

Create a custom GPT using a GPT builder tool (e.g., the GPT builder in ChatGPT Plus). In ChatGPT, navigate towards "GPTs" tab.

Step 2: Define Categories

Provide the GPT builder with a list of expense categories relevant to your needs.

Step 3: Define Categorization Logic

Provide rules or logic for assigning transactions to categories. For example: "Transactions with 'Uber' or 'Lyft' should be categorized as 'Travel.'"

Step 4: Name and Brand GPT

Give the custom GPT a name (e.g., "Expense Categorizer") and optionally, a logo.

Step 5: Data Input

Input the transaction data (from the cleaned credit card statement) into the custom GPT. This might involve copying and pasting the data.

You can also specify the capabilities of the GPT like enabling web search for it to search the web.

Step 6: Specify Model

You can customize which model your GPT will run on. This way you can choose a model fit for categorizing or maybe creating images.

I suggest using o3 for quality (reasoning) or if you prefer speed, use 4o.

Step 7: Refine Categorization (Optional)

Review the initial categorization, paying close attention to transactions categorized as "Other" or "Miscellaneous."

• Provide feedback to the GPT to improve accuracy. For example: "Review the transactions in 'Other.' Transactions with 'Office Depot' should be categorized as 'Supplies.'"
• The GPT will then refine the categorization.

The transaction data is categorized, significantly reducing the manual effort required.

Have you already started experimenting with Custom GPTs for finance tasks, maybe without realizing how far you could take it?

I’d love to hear what’s worked for you so far, and how I can support your progress in future editions.

Looking forward to your feedback.

​

Best,
Your AI Finance Coach, Nicolas

​

P.S. Come to my next masterclass and learn real-world finance use cases step-by-step.

👉 Claim your spot before all seats are filled​