How to Save Real Money on AI costs

To all my city and senior leader friends out there: AI costs are real costs. And we often struggle to measure the actual ROI of AI deployment.

If your team starts using top-tier models like Opus left, right, and center, it’s going to cost you a pretty penny. That doesn’t mean you shouldn’t use them—but you should be strategic.

Use Loops, Not One-Offs

Instead of throwing a powerful (and expensive) model at every single task, take advantage of what I call “loops.” This is essentially running a detailed spec in a tight loop until the entire task is done.

A good strategy:

• Use a higher-level model to automatically write a very detailed spec for the task.
• Then have a cheaper model—something like Haiku—actually execute the work based on that spec.

This approach can save you a ton of money on AI usage while still getting high-quality results.

Try it out in your workflows, track the difference in spend, and you’ll see the impact. You can thank me later.