In our previous article, we provided an overview of the 6 common GenAI "spend traps" we see repeatedly, and highlighted that there are savings of 20-80% available.

Following that post (see here), the number of people requesting I follow up on the intro's promise of going a bit deeper is almost at single figures.

So, thanks to the pressures of that demand, I wanted to share more thoughts on Spend Trap #1: using premium models as the default. Whilst we wouldn't usually address this in isolation, I'll keep the article focused on that.

This is one of the most impactful elements to address and, done correctly, also low risk. For reasons that will become clear, this alone can easily reduce LLM costs by up 80% in many cases.