Most business owners have heard the word "framework" thrown around a lot lately and filed it under "probably technical, not my problem." In this episode, we make the case that it is your problem, not because you need to build one, but because understanding what a framework actually does is what helps you evaluate any agent tool being pitched to you, spot where an agentic workflow is likely to break down, and make smarter decisions about what to hand over and what to keep human.Most business owners have heard the word "framework" thrown around a lot lately and filed it under "probably technical, not my problem." In this episode, we make the case that it is your problem — not because you need to build one, but because understanding what a framework actually does is what helps you evaluate any agent tool being pitched to you, spot where an agentic workflow is likely to break down, and make smarter decisions about what to hand over and what to keep human.

In this episode, Jess and Kyle walk through a complete practical example, that shows how you move from writing down your decision logic to deploying a real working agent, step by step.

We also cover two major stories coming out from Anthropic, plus, a look at Meta's first model from its Superintelligence Labs — and what it says about what big spending can and can't buy in AI right now.

What You'll Learn

What an AI agent framework actually is
The five components every agent system needs to work — model access, tools, memory, coordination, and human oversight — and why these are your map for evaluating any AI product
Why decision logic mapping is important before you build anything
How to automate a real business process, using inbound enquiry handling as a worked example, from writing your decision logic through to rolling out with guardrails
The difference between CrewAI, LangChain, and LangGraph, and which situations each one is suited to
Inbound enquiry automation as a practical use case
Human-in-the-loop and why it matters as agents gain more access

Timestamps:

00:00 Introduction and Personal Updates

06:42 Recapping AI Agents and the Ladder of Autonomy

12:07 Frameworks 101: The Five Capabilities Every Agent Needs

18:41 How Framework Design Affects Performance and Cost

19:20 Frameworks vs. Capabilities in Agent Systems

25:40 How to Get Started with Frameworks: A Five-Stage Path

40:10 Real-World Example: Automating Inbound Sales Enquiries

47:52 Safety Considerations in Agent Use

50:21 Key Takeaways for Agent Frameworks

52:28 AI News of the Week: Claude Mythos and Third-Party Harnesses

58:42 AI Gone Wrong : Meta's Expensive Model Mistake

AI Agent Frameworks Explained: The Five Things Every Agent System Needs