AI tools like ChatGPT and Gemini can sound confident — even when they’re wrong. One time, ChatGPT told me July 1, 2025 was a Monday (it was actually a Tuesday). These so-called “hallucinations” can be far more subtle in finance, from misquoting accounting rules to fabricating data. For CFOs, the takeaway is clear: AI outputs are a useful draft — but they always require verification.

Elkar, a new AI add-on for Excel, exceeded my expectations where other AI tools fell short. From handling consolidations correctly to reformatting a 200+ row P&L, it showed real accounting nuance. Elkar proves how AI can save finance teams time inside their most familiar tool.

A real-world EASY test of AI tools for gathering public company financials reveals major accuracy issues. Perplexity and Gemini both failed to produce reliable data, while ChatGPT performed best. A cautionary look at how generative AI handles earnings reports and SEC filings.

Getting bank data into NetSuite can be a tedious task for finance teams—especially with international statements. In this test, ChatGPT showed real promise by successfully converting U.S. and Israeli bank statements into ERP-ready format, outperforming Gemini, which refused to engage. A glimpse into how AI could ease a very manual part of finance.

I tested ChatGPT's ability to perform a basic accounting consolidation. Despite confident promises and reasonable initial questions, the AI failed to eliminate subsidiary equity and investment accounts - fundamental consolidation requirements.

This post explores how ChatGPT and a specialized accounting agent handle increasingly complex US GAAP questions—from basic revenue recognition to nuanced bill-and-hold scenarios—highlighting differences in accuracy, source interpretation, and the importance of "trust but verify" when using AI in financial analysis.