RAG vs. Tool Calling: How Agents Access Up-to-date Information

Jacob Bank has been building AI agents since before they were cool. While the rest of us were just figuring out what ChatGPT could do, Jacob was already 13 years deep into agent research. Now, as founder and CEO of Relay.app, Jacob is making AI agents accessible to everyone, from customer success teams to businesses that install slip-resistant floors.

Jacob has a fascinating prediction: within a year, agents will be able to handle virtually any task you do on your computer. The real question won't be "can an agent do this?" but rather "how do I teach, manage, and trust my agent?"

Timestamps

0:00 - Preview & Introduction

1:48 - Meet Jacob & An Overview of Relay.app

4:08 - From Stanford AI Research to Google to Founder

4:57 - AI Agents were 15 Years in the Making

8:30 - It Took 2.5 Years to Get the First Customer

12:35 - What Actually Is Agentic AI?

14:55 - Understanding AI Agents

19:45 - How AI Accesses Information: Context Windows, and RAG

24:11 - The Shift from RAG to MCP

25:51 - Understanding MCP: The Restaurant Analogy

30:04 - Real-World Use Cases of Workflows with AI for CSMs

33:51 - When and When Not to Use Agents

37:45 - Why SMBs Move Faster on AI Than Big Enterprises

42:12 - The Two Limitations of Agents Today

44:31 - We're All Managers Now

46:15 - Where to Stay Current on AI

47:36 - The Value of Hands-On Experience with AI Tools

What you'll learn:

- The critical difference between workflows and agents (and why you need both)

- Why "we're all managers now"—even if you're an individual contributor

- The future of work isn't about doing tasks; it's about managing and calibrating your agents

- Practical use cases of Agentic AI for CSMs

- Why spending an hour daily experimenting with AI tools is "worth its weight in gold"

Resources mentioned:

- Ben's Bytes newsletter

- Jeff Siu (YouTube)