I sit down with Fabien Cros, who runs the AI practice at Ducker Carlisle, to talk about how he is seeing decentralized AI models inside organizations work much better than a top-down, centralized approach. He has found the key is giving employees a platform to build their own agents, letting the best use cases bubble up, and using a central engineering team for creating resiliency and managing risk.

We get into why companies keep confusing researchers with engineers, why cheap deterministic code often beats an expensive frontier-model call, and why 200 people experimenting daily will always out-build 5 experts.

Enjoy.

Recording date: June 9, 2026

Chapters

00:00 Introduction

04:44 Who Is Ducker Carlisle?

06:13 What Is Decentralized AI?

09:58 Example Of Centralized vs Decentralized AI

14:46 How The Model Works Across Multiple Clients

17:18 The AI Talent Gap

24:07 Infuse With AI or Start From Scratch?

28:30 How Ducker Carlisle Works With Clients

32:31 Hiring Interns to Build AI Systems Doesn’t Work

38:58 Training Employees To Build With AI

43:50 Fabien's Background

45:54 Future Of Consulting

49:31 Choosing The Best LLMs

54:19 US vs European Market

57:57 Decentralization Is Up To Us

01:01:31 How Do You Incentivize Internal Builders?

01:03:21 Rapid Fire

Connect with Jordan:

* Follow Jordan on X: https://x.com/jrwolfe

Connect with Fabien Cros:

* Ducker Carlisle: https://www.duckercarlisle.com/

* LinkedIn: https://www.linkedin.com/in/fabien-cros/

Remember to go direct!

Jordan

P.S. Build with love

P.P.S Be an owner

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit goingdirect.substack.com