Your Agent Is Wasting Tokens and You Don't Know It - Erik Hanchett, AWS

Summary

This presentation outlines five strategies for reducing token costs when building agents, focusing on practical techniques like caching system and tool prompts, and intelligently routing requests to different models based on complexity, such as using Claude Haiku for simple tasks and Claude Sonnet for more difficult ones. The key takeaway is to avoid using the most expensive models for every operation and to offload large tool results to save on repeated token consumption.