Question

A manufacturing company wants an agent that runs on ruggedized tablets used by field inspectors in remote plants with intermittent or no internet connectivity. The agent must classify equipment defects from short text notes and return guidance within a second, even offline. Data sensitivity rules also prohibit sending inspection notes to external cloud endpoints. As the solution architect, which approach best fits these constraints?

Accepted Answer

Deploy a fine-tuned small language model (SLM) locally on the tablets for on-device inference

Answer

Extend Microsoft 365 Copilot with a declarative agent that calls a large frontier model in Microsoft Foundry

Answer

Build a Copilot Studio topic that queries a cloud-hosted GPT model through a connector

Answer

Implement a model router in Microsoft Foundry that balances between GPT-4o and GPT-4o mini