Service
App development, AI integration
Hutsy AI
About Hutsy AI
Hutsy AI is an extension of the original Hutsy Financial application, designed to enhance the user experience by providing personalized financial assistance through advanced AI capabilities. After the successful launch of the Hutsy Financial app—an ambitious project led by Tefari Bailey, CEO of Hutsy Financial—DETL Inc. continued its partnership with Hutsy to build on that momentum.
The Hutsy Financial app, developed to support over 50,000 users at launch, integrated with the Hutsy Debit Card and established a seamless payment ecosystem. Building on that foundation, Hutsy AI introduces a new dimension to the platform by implementing a custom Retrieval-Augmented Generation (RAG) model. This AI-powered chatbot offers users tailored financial insights, answering their queries based on their individual financial data. Hutsy AI expands the capabilities of the original Hutsy platform, empowering users to make more informed financial decisions through real-time, personalized support—further advancing Hutsy's mission of financial empowerment.
The Description
The Hutsy AI app required the implementation of several advanced features to deliver personalized financial assistance and enhance the user experience. The following features were essential:
- Real-Time Communication for Instant responses powered by a custom RAG model to access users' financial data securely.
- A chat-based interface designed to simulate conversations with a personalized financial advisor.
- Ability for users to set budgets and, based on transaction data, receive real-time updates on how much they have left to spend.
Owner
Tefari Bailey
Project Date
Start: 04/2024 - End: 08/2024
The Challenge
Building the Hutsy AI application presented a range of complex challenges that required careful planning and innovative problem-solving. One primary challenge was handling user financial data securely. Since we needed to fetch users’ financial information for analysis, it was critical to strip out any personal identifiers before feeding the data into the AI. This required designing a process to anonymize and compile financial data in a dedicated document, which would then be accessible to the AI without compromising user privacy. Adding to this complexity, we needed to enable users to create and adjust budgets. These budgets had to sync with the user's backend document—where the AI could access and interpret it—so the AI could provide real-time updates and insights on the user’s spending, remaining budget, and financial habits. Fine-tuning the AI to perform financial calculations accurately required us to teach the model how to manage budgeting scenarios. By training it with sample budget calculations, we ensured the AI would deliver more reliable financial advice tailored to users’ specific goals.
Another major hurdle was developing a custom chat interface that created a seamless, intuitive experience for users to interact with Hutsy AI as though it were a human financial advisor. This interface had to facilitate two-way communication while preserving the look and feel of a personalized chat. In addition, we built functionality allowing the AI to reference past interactions to create a more meaningful and personalized experience, particularly with recurring advice or questions, which brought additional technical demands around data storage and retrieval. Another challenge was incorporating credit score insights into the application. Users needed to view their credit score, while the AI needed to provide actionable feedback based on the user’s account data. Developing these recommendations involved refining the AI’s access to historical financial data and transaction records, ensuring it could offer advice on improving financial health.
The Solution
To securely integrate users’ financial data into Hutsy AI, we partnered with a reliable third-party vendor that provided controlled access to users’ financial information. Users could authenticate directly through the vendor’s login page, enabling a secure transmission of financial data via an API endpoint. This approach safeguarded user credentials while giving our system access to essential financial details. To optimize data accessibility, we compiled the financial information in a structured JSON document, allowing the AI to analyze the data while ensuring it remained organized and ready for quick access. When users created budgets, each new entry was automatically appended to this JSON document, creating a comprehensive, up-to-date financial profile. This single document housed all relevant financial information, allowing the AI to retrieve data seamlessly and support user interactions in real time.
For user privacy, we implemented stringent data-sanitization processes to strip out any potentially sensitive information, such as names or locations, to maintain anonymity without compromising the AI’s ability to offer personalized insights. We stored previous chat interactions in a separate document, allowing the AI to access historical conversations and tailor responses based on past user queries, creating a sense of continuity. Since real-time financial updates were unavailable through the vendor, we developed a scheduled background API that ran daily at 6 a.m., ensuring that the AI had the latest data without requiring constant polling. For credit score functionality, we integrated with a separate third-party credit union vendor to fetch and present users’ credit scores, ensuring that personal information was stripped away to maintain secure, anonymous AI interactions. This setup allowed the AI to offer a personalized yet anonymous experience, fostering trust and security for users as they received tailored financial insights without compromising their privacy.