Digits Blog

News and all things business finance

Bar chart comparing transaction categorization accuracy. Digits AGL® leads at 97.8%, followed by OpenAI GPT-6 Luna at 79.5%, Google Gemini 3.8 Flash at 79.2%, human outsourced accountants at 79.1%, OpenAI GPT-6.1 Sol at 78.9%, Anthropic Claude Fable 5.1 at 78.2%, Anthropic Claude Opus 5.5 at 77.9%, and TypeSafe Jev at 61.4%.
News Breaking: AI Can’t Close Your Books From the Outside (Yes, Including Jev)

Frontier models haven’t improved at accounting on their own. One-shot LLM classification has plateaued (and actually declined since June). Accuracy improves with access to ledger history, but at significantly higher cost and token usage. Ledger-native accounting models like Digits AGL® still outperform frontier models.

News Digits Adds 11 New Quality Checks

Digits adds eleven new Quality Control checks to help firms catch balance sheet issues, coding inconsistencies, uncleared accounts, and more. The checks run continuously and deterministically, surfacing exceptions for accountants to review.

Switch to Digits today

Experience accounting, reimagined.