Don’t follow the herd on AI cost optimization – control compute this way (www.techradar.com)

Recent discussions in the AI/ML community emphasize the shifting focus from merely tracking token usage to optimizing AI costs through better model governance and integration of business logic. With companies like Uber facing significant budget overruns on AI deployments, the idea of model routing and leveraging cost-efficient open-source models has gained traction, especially with new Chinese models like Kimi K3. However, experts warn that isolating model selection without a broader strategy can lead to unnecessary token consumption and inflated costs. The article highlights the importance of incorporating a business logic layer into AI systems, which enables the reuse of organizational knowledge and analytics workflows. This approach not only minimizes repeated compute efforts from large language models (LLMs) for routine queries but also ensures that the outputs are aligned with internal business definitions and compliance standards. As organizations increasingly deploy AI agents, embedding robust business logic into AI processes becomes essential for maintaining accuracy and trust, ultimately leading to better cost efficiencies and ROI in AI initiatives.