What if the biggest AI breakthrough isn't a bigger model? What if it's choosing the right model for the right workflow? One point really stood out to me. Openness may become far more important than many people realize. John also raised an important economic question. Frontier models remain expensive to operate. ServiceNow has shared that it processes roughly 80 billion workflows annually. Running every workflow on a frontier model would create an enormous inference bill. But do all of those workflows actually require frontier-level reasoning? Many could likely be handled by smaller language models that deliver the required accuracy at a fraction of the cost. That shifts the conversation from model capability to orchestration, routing, and economics. Perhaps the future is less about finding the smartest model and more about knowing when not to use it. After all, even enterprise budgets appreciate a little efficiency. How are you thinking about model selection inside enterprise workflows? Are you seeing organizations successfully balance capability, latency, and inference cost? Watch the full video: https://www.youtube.com/watch?v=yrQAVO7nrDwD Background Soundtrack: Away From You – Mauro Somm #ProjectManagement #CIO #CFO - @elevatiq