Question

A retail company is building a custom agent in Microsoft Foundry that handles a mix of requests. Roughly 70% are simple FAQ lookups and short summaries, while 30% require complex multi-step reasoning over product catalogs and financial data. Leadership wants to minimize total cost of ownership without sacrificing quality on the complex requests. Which approach best meets these requirements?

Accepted Answer

Implement a model router that directs simple requests to a small/efficient model and complex requests to a more capable large model based on request characteristics.

Answer

Route every request to the most capable large reasoning model to guarantee consistent quality across all request types.

Answer

Route every request to a single small language model and add more retrieval grounding to compensate for reasoning limitations.

Answer

Deploy two entirely separate agents and require users to manually choose which one to use for each request.