AI Product Development
A mid-sized Ohio city struggled with an outdated tax processing system that required municipal employees to manually enter data from complex tax forms, leading to significant delays in return processing. Because the manual data entry and physical letter notifications to taxpayers were slow and labor-intensive, the city was missing quarterly revenue collection deadlines.
The city required a modern, automated solution to extract data from scanned tax forms, apply business rules to calculate taxes or refunds, and ensure accuracy, all while remaining secure and accessible only to municipal employees. The new system also needed to handle both standardized and non-standardized forms, with plans for future e-filing and data migration from the old system.
Taazaa developed an AI tax filing assistant to streamline tax processing for the city, leveraging a robust tech stack.
The back-end, built with Python and Java, used PostgreSQL and MongoDB for data storage, with Apache Kafka enabling seamless inter-component communication. The front-end, developed using Next.js with TypeScript, incorporated Tailwind CSS and Ant Design for styling, Vitest for unit testing, and React Query for efficient API calls. ESLint, Prettier, Lint Staged, and Husky ensured code quality.
For the AI, the system integrated Microsoft Azure’s Document AI and Google’s Gemini, alongside Python and JSON, to scan and extract data from PDF tax forms, storing values in a database. The AI extraction achieved up to 80-85% accuracy, and an edit functionality was implemented to allow municipal auditors to manually correct or input missing values.
The platform was designed exclusively for municipal employees and was not accessible to the public. It featured customized data models to read any PDF form, including non-standard text or value formats. Middleware ensured data integrity, and business rules computed taxes or refunds.
The first phase, including AI extraction and edit functionality, was delivered successfully. Future phases are planned for an e-filing system and data migration from the old system using UiPath RPA tools, enabling scalability for other U.S. municipalities.
Taazaa’s AI Tax Filing Assistant enabled municipal officers to meet revenue collection targets with greater efficiency. By automating data extraction from tax forms, the system eliminated manual entry, reducing processing times from days to hours.
Tasks that previously delayed quarterly deadlines were streamlined, allowing the city to process more returns within target periods and boost revenue inflow.
The edit functionality ensured accuracy by allowing auditors to address the 80-85% AI extraction accuracy swiftly. Powered by a sophisticated tech stack, the platform can handle both standardized and non-standard forms. Due to other applications lacking API support, the Tax Filing application uses RPA to export the data into another city system.
With planned e-filing, the solution promises further efficiency gains, earning trust from municipal stakeholders through reliable delivery and setting the stage for broader adoption across U.S. municipalities.
.svg)
