How to build Enterprise Aware Agents
summary
Argues for a practical Workflow Search paradigm that turns agent traces into reusable, evaluable workflows and backfills enterprise context into LLM loops. Describes using a library of golden workflows for evaluation and training (fine-tuning or RLHF) and an alternative of dynamic prompting that retrieves similar workflows at runtime via hybrid search + rerank, augmented by knowledge‑graph authoritativeness signals. Demonstrates agents generating workflows, workflows training agents, and tradeoffs: fine‑tuning for scale, search for fast adaptability.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:52Workflows
- 2:14Agents
- 2:59tradeoffs
- 6:08workflow evaluation
- 7:10workflow training
- 8:00workflow generation
- 8:44workflow discovery
- 9:20thought experiment
- 11:18finetuning
- 13:18dynamic prompting
- 14:20competitor analysis example
- 15:04comparison time
- 18:18key takeaways
related talks
14mOrchestrating Agentic Workflows with Governance and Trust
25mFrom Chaos to Control: Building Enterprise-Grade LLM Agents That Actually Work
35mAgentic Workflows: Stories from the Engine Room
49mAI Agents: Collaboration in Real-Life Applications
22mAgentic AI for Engineering Leaders | Strategy & Execution Guide
14mUsing Graphs to Fine-Tune Agents
47mHow to Build Enterprise-Grade AI Agents Using Robust Evaluation
22mOrchestrating Multimodal RAG With Agentic Workflows With OpenSearch