Using Chat GPT to Build a 3 Way Model Whilst Keeping Your Data Secure and Private

As the field of finance continues to embrace the advancements in artificial intelligence (AI), the integration of AI models, such as ChatGPT, holds immense potential for financial planning and analysis (FP&A) professionals.

In particular, the Code Interpreter feature within ChatGPT offers exciting opportunities for FP&A tasks.

While some may believe that adding private data is necessary for generating accurate models, it is indeed possible to create a 3-way model using ChatGPT without compromising sensitive information.

This article aims to provide FP&A professionals with a step-by-step guide on how to leverage ChatGPT to develop a 3-way model while maintaining data privacy.

FULL GUIDE

Step 1: Enable Code Interpreter

Acquire ChatGPT+ and Enable Code Interpreter: To begin, obtain ChatGPT+ and ensure that the Code Interpreter functionality is enabled.

This feature allows ChatGPT to interpret and generate Python code, making it a powerful tool for building financial models.

Step 2: Identify the Required Data:

Engage with ChatGPT and inquire about the specific data it requires for model development. By understanding the necessary inputs, you can effectively prepare dummy data that will serve as a placeholder during the initial modeling phase.

For example, to build a 3 way model, Chat GPT asked for several inputs, in the form of Excel files:

• Historical Financial Statements: This includes historical data for the Income Statement, Balance Sheet, and Cash Flow Statement for the past 3–5 years or more. This is used as a basis to forecast future financials and understand the historical trends.
• Sales/Revenue Forecasts: Forecasts for the company’s sales/revenue for the future period that the financial model will cover. This could be broken down by product, region, or any other relevant segmentation.
• Expense Forecasts: Forecasts for the company’s expenses. This could include cost of goods sold (COGS), operating expenses, capital expenditures, and other expenses. Like the revenue forecast, this could also be broken down into more granular details.
• Capital Structure Details: This includes the company’s existing debt schedule, any future debt or equity issuances, and the cost of debt and equity.
• Tax Rates: The company’s effective tax rate or the tax rates in the jurisdictions in which it operates.
• Working Capital Assumptions: Assumptions around how the company’s working capital (accounts receivable, inventory, and accounts payable) will evolve over the forecast period.
• Depreciation & Amortization Schedules: This includes details on the company’s existing fixed assets, their useful lives, and any future capital expenditures.
• Investment Plans: Any planned investments in new projects or business lines, including expected costs and incremental revenues and profits.
• Other Assumptions: Other relevant assumptions that could impact the company’s future financial performance. This could include macroeconomic factors, changes in regulatory environment, etc.

It’s important to note that the exact inputs will depend on the company’s specific situation and the level of detail required in the financial model.

Step 3: Feed Dummy Data and Run the Model:

Supply the dummy data to ChatGPT and initiate the modeling process. ChatGPT’s Code Interpreter feature will generate Python code based on the provided data, creating a foundation for the 3-way model.

Step 4: Access the Generated Python Code:

Navigate to the “workings” section within ChatGPT to obtain the Python code that has been generated. This code represents the logic and structure of the model, acting as a blueprint for further development.

Step 5: Paste the Code into Your Python Environment:

Copy the Python code from ChatGPT and paste it into your preferred Python environment, such as Google Colab or Microsoft Azure. These platforms provide a user-friendly interface for executing Python code and offer collaborative features to facilitate team collaboration.

If you want to get started with Google Colab and Python, read this:

Step 6: Incorporate Real Data:

Replace the dummy data in the code with actual financial data relevant to your FP&A analysis. This step allows you to leverage the power of ChatGPT’s generated code while maintaining data privacy, as the private data is never exposed to the model-building process.

Step 7: Refine and Enhance the Model:

With the real data integrated into the Python code, you can begin refining and enhancing the 3-way model according to your specific FP&A requirements. This may involve incorporating additional calculations, performing sensitivity analyses, or incorporating advanced statistical techniques.

For this I used this prompt:

Now I want you to do 5 things: 1. Tell me some insights of the data 2. Create some visualizations with this data 3. Summarise your findings in 5 bullet points for a CFO 4. Build a forecast for the next 3 years for each account. Output in Excel 5. Suggest me next steps in terms of more financial analysis and data analysis

So you can get data visualizations.

Or a summary for your CFO.

Or a financial forecast.

And even suggestions on what are the next steps.