AI Models Get 57% of Financial Advice Wrong

Saturn tested modern AI models in the role of financial advisers, and the result was sobering: 57% of the answers contained errors. The models delivered money advice confidently and without caveats, even when the steps they suggested were wrong.

The issue isn't that AI makes mistakes — people do too. The problem is that a wrong answer sounds just as smooth and convincing as a correct one, and spotting the difference without an expert review is nearly impossible. When the mistake affects company money, that gap becomes costly.

Check who in your company already asks a chatbot about taxes, loans or investments and acts on the answer without a second look. The approach differs here: 1Cifer's agents answer with figures pulled from the company's 1C, bank and CRM data, and payments or approvals move through a routed process with a decision history — not on the strength of one model's reply.