In our previous article on 6 GenAI spend traps, we covered the common ways organisations leak money on GenAI, including using premium models as the default (which got its own deeper-dive).

This article focuses on Spend Trap #2: using GenAI where rules or traditional ML are better suited.

This is one of those problems that sounds obvious when stated, but in practice is remarkably common. You don't need a sledgehammer/1,000 ton hydraulic press to crack a nut.

When you've got a shiny new hammer (Opus 4.5, say), everything starts looking like a nail. And when that hammer costs £20 per million tokens, those nails get expensive fast (assuming the dollar doesn't completely collapse...).