More accurate tracking of AI costs and tasks
We fixed how our system tracks and labels AI work behind the scenes, giving us a clearer picture of what these features actually cost.
Behind the scenes, our AI features rely on several different tasks — like generating content for different audiences, running quality checks, and creating shipping summaries. We found that these tasks were sometimes being mislabeled in our internal cost tracking, which made it hard to know exactly what was driving expenses.
We corrected this labeling so each type of AI task is now properly categorized. We also added more detailed pricing tracking for each AI model we use, along with improved logging of AI conversations. Together, these changes give us a much clearer and more reliable view of how our AI systems are being used and what they cost.
With accurate cost tracking in place, we can better monitor and optimize how we use AI behind the scenes. This helps us keep the service running efficiently and make smarter decisions about AI usage going forward.