Definition
Product architecture that dynamically or statically assigns user prompts to different underlying LLMs (in-house vs third-party, cheap vs frontier) based on task complexity, cost, latency, or policy.
Key Points
- microsoft Excel/Outlook MAI routing (Jul 2026) is a high-volume cost example
- Often opaque to end users (no model disclosure in Copilot UI)
- Enables cheaper-default / premium-frontier pricing tiers
Related
- microsoft-mai-replaces-openai-anthropic-office
- llm-pricing
- model-orchestration
- mai-thinking-1
- vendor-lock-in