Building Agentic Workflows: Patterns for Orchestrating Intelligent Systems.
summary
Claim: orchestrating multiple specialized agents yields reliable product automation rather than one monolithic AI. Implementation shows a Postman “agent mode” prototype with three specialized agents (knowledge agent using react reasoning, coding agent using plan-and-execute, resource agent using tool-call loops) coordinated by a manager-worker orchestrator that generates, validates and injects a contract test into a request. Tools include Postman AI Request comparisons, KB queries via an MCP server, and models like GPT‑4-mini and GPT‑5-mini (GPT‑5 ran ~17.7s vs GPT‑4 ~750s in one comparison).
AI-written summary · about
chapters (9)
- 0:00Agentic workflows intro
- 1:52Core agentic components
- 4:03Agent perception memory context
- 5:28Single agent orchestration
- 8:34Multi-agent coordination
- 11:48Choosing orchestration patterns
- 12:55Designing AI agent workflow
- 17:31Selecting models for agents
- 24:16Recap and key takeaways
related talks
40mBuilding Effective AI Agents Simple, Composable Patterns
25mAI Orchestration: Getting Super Agents to Complete a Mission
27mEmerging Patterns Agentic AI
27mMulti-Agent AI: Transforming Business Operations and Value Creation
29mWhat exactly are AI Agents?
29mPairing with the machine: How AI Agents are rewriting the rules of software development
34mAI Agents: Transforming Autonomous Decision-Making
33mAsk Me Anything: AI Agents, Skills, Career & Growth