Build sophisticated custom copilots with Azure OpenAI Assistants
summary
Azure OpenAI announces Assistants as a stateful agent platform with concrete new features: File search (RAG-as-API, supports up to 10,000 files) that extracts, embeds and stores content in Azure AI Search; customer‑managed key encryption for prompts and files; prompt shields to block injection; streaming responses; code interpreter for Python analytics and visuals; function calling (parallel and secure via managed identity); browse via Bing; and AutoGen integration for multi‑agent workflows. Claims: 6,000+ customers building, 12,000 customers and 700,000 assistants in three months, and multi‑agent setups that push task correctness from ~67% to >95%.
AI-written summary · about
chapters (14)
- 0:00Introduction to AI agents
- 0:35The evolution of AI apps
- 3:44Benefits of the assistants API
- 5:58New features at Build
- 8:28File search demo
- 14:07Code interpreter demo
- 17:29Advanced data forecasting
- 20:48Security and prompt shields
- 22:13Browsing tool demo
- 24:15Function calling demo
- 27:39Multi-agent workflows
- 32:04Freshworks customer case study
- 39:35Coca-Cola customer case study
- 45:46Closing remarks
related talks
15mAI Studio - Creating and scaling your custom copilots
45mThe era of the AI Copilot
48mGetting started with generative AI using Azure OpenAI Service
58mNext generation AI for developers with the Cloud
10mAI Co-Pilot Demo
41mGoing beyond copilot with AI Agents
37mHarnessing OpenAI to help millions
13mAI from Microsoft's Perspective