Code for America will use AI to screen tax documents for volunteer preparers
A Google.org grant funds a tool that checks uploads for GetYourRefund, which has helped 387,000 low-income Americans claim $657 million.
Code for America will expand an AI tool that helps volunteers process tax documents for low-income families, using funding from the $30 million Google.org Impact Challenge: AI for Government Innovation, StateScoop reported. The non-profit is one of 15 recipients, a programme IQGov [covered when it was announced](/article/google-org-ai-government-challenge/).
The tool supports GetYourRefund, Code for America's virtual tax-filing service, launched in 2020 with IRS support and Volunteer Income Tax Assistance (VITA) organisations. It has helped more than 387,000 Americans file and claim nearly $657 million in refunds.
Routine screening, human judgment
The AI will review documents taxpayers upload and flag which cases can move straight to preparation and which need more review. The aim is to let volunteer preparers spend less time on routine checks and more on complicated questions.
"We believe in AI that makes people's lives better," said Amanda Renteria, Code for America's chief executive. For families using the service, she said, it means "an easier path to filing their taxes, claiming the benefits they have earned".
After Direct File
The project comes as free filing options have narrowed. The federal government discontinued Direct File, its free service for filing directly with the IRS, last year. In July Code for America began testing a platform called Simple File with Colorado and New Jersey.
Tax credits such as the Earned Income Tax Credit are among the largest anti-poverty programmes in the US, but they only reach people who file. Volunteer capacity is the bottleneck in free tax help, so a tool that triages paperwork is a sensible use of AI: it speeds up the queue without taking decisions about people's taxes away from trained preparers.
Newsroom, Online
Stories by the IQGov newsroom are researched and written by AI agents to our editorial rulebook, checked against their sources by a separate fact-checking agent, and approved by the publisher before publication. Every fact links to its source.
