DataFramed #316 Enterprise AI Agents with Jun Qian, VP of Generative AI Services at Oracle
63 snips
Aug 18, 2025 Jun Qian, Vice President of Generative AI Services at Oracle, dives into the transformative potential of AI agents and chatbots. He discusses the evolution from traditional chatbots to advanced systems that utilize large language models and enterprise knowledge. Exploring the challenges of managing knowledge bases, Jun emphasizes the need for effective governance before deployment. He also addresses metrics for measuring AI effectiveness and future innovations in coding and robotics, highlighting Oracle's ambitious Stargate project.
AI Snips
Chapters
Books
Transcript
Episode notes
Why ChatGPT Feels Like An Agent
- ChatGPT stands out because it combines model access with memory and tool use, making it feel like an agent.
- Jun predicts future GPT versions will behave more like full agents than standalone models.
Memory And Tool Use Are Core Agent Features
- Memory and tool calling are core agent capabilities that make interactions contextual and persistent.
- Jun points out ChatGPT's memory feature as a major factor in its usefulness.
Combine LLMs With RAG, Not Replace Knowledge
- Use large language models to remove brittle intent classifiers and enable natural follow-ups.
- Keep a retrieval layer (RAG) for up-to-date facts and maintain knowledge-refresh practices.
