AI coding assistants went from “nice-to-have” to “this how we work”. Then the vendors started changing prices...

As usage-based billing models become standard, engineering leaders are put in an uncomfortable position. The tool your teams rely on starts behaving like a variable cloud bill. Costs rise quietly, finance starts asking difficult questions, and suddenly you are forced into a false choice. Either absorb the additional costs, or restrict usage and lose the momentum.

This session is about finding the third option: keeping the productivity gains while making spend more predictable and under control. Based on our organization's experience, we'll show you how we coped with changes to GitHub Copilot's pricing model, and even reduced our overall AI tooling costs. We will break down what actually drives AI costs in development and how to influence each lever. From there, we will walk through three high-impact techniques to dramatically reduce token and model usage without sacrificing productivity, velocity, output quality, or developer experience.

Rated level 200, this talk is designed for engineers, tech leads, and engineering managers who want to scale AI assistants responsibly. The insights shared will be valuable for both technical and non-technical professionals, as techniques may apply to any AI-assisted knowledge work, not just software engineering.