Intuit Slashes Tax Code Work from Months to Hours

Intuit’s latest internal hack turned a task that usually drags on for months into something that can be finished in a handful of hours. The company says the breakthrough came from re‑thinking how tax‑code logic is written, not from adding more engineers or throwing raw computing power at the problem. Instead, developers built a workflow that forces every change—whether a new provision or an amendment to an existing rule—to pass through a single, tightly controlled codebase.

That approach lets a regulated‑industry team keep a clear audit trail while still moving fast enough to meet filing deadlines. But the real test, according to the engineers, is whether an AI model can actually understand and extend that code. The challenge isn’t just parsing legal language; it’s translating it into a syntax that the model has never seen and making sure new clauses play nicely with decades of legacy logic.

That tension sits at the heart of Intuit’s claim that its tax calculation engine is built on a proprietary domain‑specific language maintained internally at Intuit. Any model generating code for that codebase has to translate legal text into syntax it was never trained on, and identify how new provisions interact with decades of existing code without b.