Dynamic model routing will follow the path blazed by software-defined wide-area networks

The planned acquisition of OpenRouter Inc. by Stripe Inc. has highlighted the importance of dynamic model routing. As artificial intelligence becomes more prevalent in businesses, organizations are realizing that a single model is not sufficient for every task. Different models have unique strengths in performance, latency, cost, privacy, and domain-specific accuracy.

Dynamic model routing offers an intelligent decision layer that directs tasks to the most suitable model or inference endpoint based on current needs and conditions. This concept is similar to the shift from static decision-making in software-defined wide-area networking (SD-WAN). As IT environments became more distributed and business-critical, SD-WANs evolved from simple connection establishment to delivering optimal experiences through application-aware, policy-driven, and adaptive systems.

Model routing introduces a similar decision-making challenge at a different layer of the stack, focusing on determining the best model, endpoint, or environment for an AI request based on various factors such as application type, path performance, security policy, availability, and business priority. Just like SD-WAN, model routing requires a robust foundation with clear policy, visibility, and controls to ensure agility rather than unpredictability.

Observability platforms provide IT teams with a consolidated view across different environments, allowing continuous validation of routing decisions. The need for intelligent dynamic routing is particularly crucial at the branch and edge, where AI-powered experiences reach employees and customers. As AI workflows span across branches, clouds, SaaS applications, data stores, and multiple AI models, each step may have different requirements, such as maintaining data within specific environments, prioritizing low latency, cost efficiency, or stronger security controls.

The branch and edge will need networking, security, automation, and observability working together to ensure secure and reliable AI service operation.