AI Explained Building Agents at Scale: Lessons from the Front Lines With Gary Stafford
Sep 25, 2025
Join Gary Stafford, Principal Solutions Architect at AWS Strands Agents, as he breaks down the world of agentic systems for enterprises. He discusses when to choose agentic approaches over traditional AI, highlighting real-world use cases like sales process automation and back-office automation. Gary also explains how agents combine LLMs, tool usage, and reasoning, and the importance of the Model Context Protocol for secure interactions. Dive into the challenges of scaling, safety measures, and the balance between no-code solutions and expert engineering.
AI Snips
Chapters
Transcript
Episode notes
Work Backwards Before Choosing Agents
- Work backwards from the business problem before choosing agentic or ML solutions.
- Prefer existing analytics or ML if they solve the need faster and cheaper than agentic tech.
Agents Are LLMs Plus Tooling And Planning
- Agents combine an LLM, tool access, and a reasoning loop to plan actions.
- That planning about which tools to use is what separates agents from single-turn GenAI calls.
Automating Sales Playbooks With Agents
- Gary described automating sales touches by teaching an agent the playbook and connecting it to CRM data.
- The agent can intelligently touch 100% of customers using institutional knowledge instead of brittle hand-coded rules.