Plan a responsible generative AI solution
Unit 2 of 8 · Module 6: Implement a responsible generative AI solution in Microsoft Foundry
Read this lesson
Four stages
Microsoft's guidance for responsible generative AI is meant to be put to work, not just read. It lays out four stages for building and running a responsible AI plan whenever you use generative models, a bit like an orchard keeper protecting a harvest. Stage one, map: list the harms that could realistically come from your particular solution, like walking the rows and pinning every sign of blight or pests. Stage two, measure: check how often those harms actually show up in what your solution produces, like sampling baskets and counting the bad apples. Stage three, mitigate: reduce the harms at several layers, like nets, pruning, and a fence, and tell users transparently about the risks that remain, like an honest sign for visitors. And stage four, manage: run the solution responsibly, with a plan for deployment and day-to-day operations that you actually follow, all season long.
A familiar framework
These four stages line up closely with the functions of the NIST AI Risk Management Framework, a widely used standard. The rest of this module walks through each stage, with practical steps you can take to build a generative AI solution that works well, and behaves responsibly.
Facts from Microsoft's documentation: Overview of Responsible AI practices for Azure OpenAI in Foundry Models (CC BY 4.0), rewritten in Loop's own words and examples. Narration, animation, and quiz by Loop. Microsoft, Azure, and Microsoft Foundry are trademarks of Microsoft. Loop is independent and not affiliated with or endorsed by Microsoft.