NewsTradingSentimentEventsCommunityBriefing
Tech

Google Funds Code for America's AI Tax Tool

By Tech Desk · · 2 min read
A flat-vector illustration of a stack of blank paper documents and a calculator on a desk
Illustration: Tradingbird

Code for America will use Google.org grant funds to deploy an AI system that screens tax documents for low-income filers.

Key points

  • Code for America is using a Google.org grant to build an AI tool that screens tax documents for low-income filers.
  • The tool aims to free up volunteer preparers to handle complex cases, building on a service that has helped 387,000 people.
  • Only three US states have made major strides in AI adoption, while most remain in early pilot stages according to a nonprofit report.

Code for America has secured a significant portion of the $30 million Google.org Impact Challenge to expand its artificial intelligence capabilities in the public sector. The funding specifically supports the development of an automated document screening tool for GetYourRefund, a virtual tax-filing service designed for low-income families. This initiative is part of a broader effort to integrate AI into government services and reduce administrative bottlenecks for vulnerable populations.

The nonprofit is one of 15 organizations selected for the AI for Government Innovation grant. By automating the initial review of uploaded tax documents, the tool aims to streamline the workflow for volunteer tax preparers. This allows human experts to focus their time on complex tax questions rather than routine data entry and document verification, potentially speeding up the process for users.

Streamlining volunteer tax preparation

GetYourRefund, launched in 2020 with IRS support, has assisted over 387,000 Americans in claiming nearly $657 million in refunds. The new AI component will review taxpayer uploads to determine if cases are ready for preparation or require further human review. This division of labor is intended to make the filing path easier for families seeking financial security and earned benefits.

Amanda Renteria, CEO of Code for America, stated that the project demonstrates the potential of AI to improve daily life. The organization views this as a practical application of technology to reduce bureaucratic friction. The goal is to ensure that eligible individuals can access their refunds more efficiently without getting stuck in administrative queues.

Context of broader digital shifts

This expansion occurs as Code for America adapts to changes in the federal tax landscape. Following the discontinuation of the IRS's Direct File program, the nonprofit partnered with Colorado and New Jersey to test Simple File, a new online filing platform. The organization has also been using AI to make government documents more accessible and helping SNAP caseworkers navigate eligibility rules.

StateScoop notes that while many states are interested in AI, adoption remains uneven. A recent report by the nonprofit found that only New Jersey, Pennsylvania, and Utah have made significant strides in building AI capabilities. The remaining 47 states are largely in the developing stage, often relying on external partnerships for technical support and pilot programs.

Trade-offs of automated screening

The primary benefit of the AI tool is efficiency, reducing the time volunteers spend on repetitive tasks. However, the reliance on automated screening introduces a layer of technological mediation between the taxpayer and the human preparer. If the AI misclassifies a document, it may delay the filing process or require additional manual intervention, potentially introducing new points of failure.

Furthermore, the effectiveness of the tool depends on the accuracy of its algorithms and the quality of the data provided by users. While the intent is to reduce administrative burden, the system must balance speed with precision to avoid errors in tax preparation. This highlights the ongoing challenge of integrating AI into sensitive financial services where mistakes can have significant consequences for users.

Based on reporting by StateScoop, compiled by the Tradingbird desk.

Read next

More in Tech

More from the Tech desk

All desk stories