Intuit's TurboTax team successfully used artificial intelligence tools to dramatically accelerate the implementation of complex tax legislation, compressing what traditionally took months into a matter of hours.

The company faced the challenge of implementing provisions from a 900-page unstructured legislative document without standardized formatting or published IRS forms, while maintaining strict accuracy requirements for consumer tax preparation software.

Joy Shaw, director of tax at Intuit with over 30 years at the company, said the team used large language models including ChatGPT for document analysis, parsing, and provision filtering. The AI tools helped reconcile differences between House and Senate versions of the legislation and identified which provisions would affect TurboTax customers.

For code implementation, the team relied primarily on Claude AI to translate legal text into Intuit's proprietary domain-specific programming language used for tax calculations. The AI could identify dependencies and interactions with existing code without breaking established functionality.

Intuit also developed two custom tools during this process: one that automatically generates TurboTax product screens from legal changes, and a specialized unit testing framework that identifies specific code segments causing failures and enables corrections within the testing environment itself.

Sarah Aerni, Intuit's VP of technology for the Consumer Group, emphasized that human tax experts remain essential for validating AI-generated output, as accuracy requirements for consumer tax products must approach 100 percent. The company trained employees across all functions to use AI tools, rather than limiting access to engineering teams.