AI And Crypto: Digital Assets Thoughts Of The Week

Why AI will select AI

“The sheer number of AI agents is already unmanageable at enterprise scale. Gartner predicted that the average Fortune 500 company will go from less than 15 AI agents in 2025 to more than 150,000 in 2028, while only 13% of companies have appropriate AI governance. At this scale, it will be impossible to manually determine what model each agent should use for each task, and this is exactly what model routers excel at.”

“Most companies are drowning in token costs, and one third have responded with ‘emergency spending freezes’ on AI. Companies need to put the best frontier models in the hands of their talented employees, but costs are becoming prohibitive. What is needed is an automated system that selects the best model for each employee-initiated task, which again requires delegating to AI itself.”

“Despite many attempts to benchmark AI models, no single winner has triumphed across the board, because each model has different strengths. Case in point, 37% of organizations are using five or more different models. Cost, latency, and accuracy vary significantly by task, which means AI models must be evaluated on a task-specific basis. Once again, only AI can perform this type of task-specific model evaluation at runtime, at any kind of scale.”