Andrew GazdeckiSerial Entrepreneur Driving Startup Acquisitions

How Do You Manage AI Costs As You Scale?

28 May 2026 · 1 min

Watch on YouTube

0:53

In this brief talk, the speaker outlines two primary strategies for controlling AI expenses as usage grows: (1) select the most appropriate model for the task rather than defaulting to the newest, most expensive option—older generations can often meet requirements at a fraction of the cost; and (2) carefully manage the amount of context (tokens) sent in prompts, because larger prompts increase API token charges. By applying these practices—right‑sizing the model and minimizing unnecessary prompt content—developers can prevent runaway AI costs when features scale.

Chapters

How do you manage AI costs as you scale? Two things come to mind. The first is make sure you're using the right model for the job. Resist the urge to always use the latest model. You may not need it, especially when these models are getting so good that a few generations of models back can a lot of times solve the problem you're trying to solve for a fraction of the cost. The second way I would say is think about what you're putting into the prompt in the context when you're making API calls to these AI companies. The more you throw in, the more tokens that's going to cost. And so if you build a feature and you're just throwing everything into AI, if that feature scales, you could end up with pretty significant AI costs. Those are probably the two things that come to mind. Use the right model and always think about the prompts you're sending to AI and make sure you're not just overloading them without getting any significant benefit back from them.