Building Trust in AI Agents
summary
Claims a pathway to making agentic development actually reliable with concrete practices: choose and compare LLMs, add RAG and a vector database for company knowledge, use human-in-the-loop checks, guardrails, memory controls and an orchestration layer. Cites a Microsoft result where the top agent simply nudged people on open pull requests, and an automation that cuts 3‑week legal document creation down to one day. Digs into failure modes (hallucination, data poisoning) and how each is detected and mitigated.
AI-written summary · about
chapters (10)
- 0:00Trust in AI agents
- 1:53AI expectations vs reality
- 4:03AI agents and core problems
- 5:54How LLMs work and limits
- 10:13AI struggles with real-world concepts
- 12:43Trustworthy AI building blocks
- 16:59AI governance and security
- 20:12Trust-building summary
- 21:38AI adoption fears and challenges
- 24:45Q&A on poisoning and reliability
related talks
33mUnlocking Agentic AI: [...]
27mHow to Build AI Agents and Keep Your Sanity
22mAgentic AI for Engineering Leaders | Strategy & Execution Guide
24mAI Governance for developers: A practical framework for deploying autonomous coding agents safely…
34mAI Agents: Transforming Autonomous Decision-Making
24mHow to Build Trustworthy AI
29mPairing with the machine: How AI Agents are rewriting the rules of software development
16mBuilding Applications with AI Agents