How I Automated My Assistant's Expenses Tracking with AI

Prior to implementing the new system, the expense tracking process was highly inefficient and time-consuming. Here's a breakdown of the old process:

• Nina would request screenshots of the weekly transactions from the CEO.
• The CEO would send the screenshots, which Nina would then have to manually process.
• Nina would use the iOS or Mac OS OCR to extract the text from the screenshots.
• She would then use a pre-written ChatGPT prompt to categorize the expenses and format the data.
• Finally, Nina would manually enter each expense into a Notion database.

This process was taking Nina a minimum of 2 hours per week to complete. The CEO was unaware of the true time investment required and thought it was a much quicker task.

The main pain points of the old system were the manual data extraction, categorization, and entry into Notion. This was a tedious and error-prone process that was significantly impacting Nina's productivity and the CEO's visibility into his expenses.

Over the past few months, I've revolutionized and streamlined the system for tracking my expenses. The key component of this build is the integration of GPT Vision, which allows for a largely automated process.

Here's how the new system works:

1. My bank sends automated SMS notifications for each transaction, which I screenshot and send to my assistant, Nina.
2. Nina uploads the screenshots to a WhatsApp chatbot I've created using Voiceflow. This chatbot leverages GPT Vision to extract the transaction details from the images.
3. The extracted data is then automatically added to a Notion database, where it is displayed in a chart for easy visualization of my monthly spending.

The new system has significantly reduced the time and effort required to track my expenses. Previously, it was taking Nina up to 2 hours per week to manually process the transactions. Now, the process is largely automated, with Nina only needing to upload the screenshots and verify the data.

The integration of GPT Vision is the key to this streamlined approach. By using a custom Voiceflow function, I'm able to easily integrate the OpenAI Vision API into my WhatsApp chatbot, allowing it to extract the necessary information from the expense screenshots.

This project has not only improved my personal finance management but also showcases how you can leverage AI technologies, such as GPT Vision, to automate various tasks. I've shared all the templates and resources used in this build in my free online community, so feel free to check them out and replicate or adapt the system for your own needs.

The key to this build is the integration of GPT Vision and WhatsApp using Voiceflow. Here's how it works:

1. The bank sends automated SMS notifications for each transaction, which I screenshot and send to my assistant Nina.
2. Nina uploads the screenshots to a WhatsApp number connected to the Voiceflow chatbot.
3. The Voiceflow chatbot uses a custom function from Flowbridge to integrate with the OpenAI Vision API and extract the transaction details from the image.
4. The extracted data is then sent to a Make.com automation, which adds each transaction as a row in my Notion expense tracker.

The Voiceflow chatbot handles the entire process, from receiving the image input to categorizing the transactions and updating the Notion database. This streamlines the expense tracking workflow and saves Nina significant time compared to the previous manual process.

The key components of this build are:

1. Voiceflow: Used to create the WhatsApp chatbot interface and integrate the GPT Vision functionality.
2. Flowbridge: Provides a custom function to easily connect Voiceflow to the OpenAI Vision API and handle non-text inputs like images.
3. Make.com: Automates the process of adding the extracted transaction data to the Notion expense tracker.

By leveraging these tools, I've been able to create a highly efficient and automated expense tracking system, reducing the time and effort required from my assistant Nina. This project is a great example of how you can use your own curiosity and problem-solving skills to build AI-powered solutions for personal or business use.

The key takeaways from this project are:

• Automating expense tracking can save significant time and effort. The new system leverages GPT Vision and integration with WhatsApp to streamline the process.
• Building custom AI solutions to solve personal problems can lead to valuable learning experiences and transferable skills.
• Following your own curiosity and "scratching your own itch" is an effective way to gain practical experience with AI and other technologies.
• The resources and templates used in this project, including the VoiceFlow template and Make.com automation, will be shared in the creator's free online community for others to utilize.
• Continuously improving the system, such as adding category-based expense tracking, can further enhance the functionality and usefulness of the solution.
• The project demonstrates how AI can be integrated into personal workflows to increase productivity and efficiency.