Orchestrating Agentic Workflows with Governance and Trust
summary
Argues enterprise agents must be hybrid, combining probabilistic LLMs with deterministic components like a scorecard model or logistic regression and business rules; shows a demo orchestration of specialized agents (complaints, churn agent, email) that run behind a single endpoint container, call external services, use session memory, produce real‑time explanations and successfully send an email report. Emphasizes guardrails, data access controls, model agnosticism and the need for end-to-end lineage and escalation rules for high‑risk decisions.
AI-written summary · about
chapters (6)
- 0:00Introduction to agentic workflows
- 1:57Challenges in enterprise AI agents
- 4:46Building blocks for AI agents
- 7:37Complaints agent example
- 9:22Orchestrating multi-agent systems
- 12:16Governance and summary
related talks
16mAgents vs Workflows: Why Not Both?
26mFrom Chaos to Choreography: Multi-Agent Orchestration Patterns That Actually Work
50mHarnessing Event-Driven and Multi-Agent Architectures for Complex Workflows in Gen...
35mAgentic Workflows: Stories from the Engine Room
20mHow to build Enterprise Aware Agents
50mAI Agentic Workflows & Design Patterns
27mAgentic Systems: Leadership, Risk, and Accountability
45mAn Agent Walks Into a Bar... and Meets Another Agent: Multi-Agent Systems in Action