95% AI Project Failures | DeepSeek vs Big Tech | Liquid AI on Mobile | Google Mango Breakthrough

Happy Friday, everyone! Hopefully you got some time to rest and recharge over the Labor Day weekend. After a much needed break, I’m back with a packed lineup of four big updates I feel are worth you attention.

First up, MIT dropped a stat that “95% of AI pilots fail.” While the headlines are misleading, the real story raises deeper questions about how companies are approaching AI. Then, I break down some major shifts in the model race, including DeepSeek 3.1 and Liquid AI’s completely new architecture. Finally, we’ll talk about Google Mango and why it could be one of the most important breakthroughs for connecting the dots across complex systems.

With that, let’s get into it.

⸻

What MIT Really Found in Its AI Report

MIT’s Media Lab released a report claiming 95% of AI pilots fail, and as you can imagine, the number spread like wildfire. But when you dig deeper, the reality is not just about the tech. Underneath the surface, there’s a lot of insights on the humans leading and managing the projects. Interestingly, general-purpose LLM pilots succeed at a much higher clip, while specialized use cases fail when leaders skip the basics. But that’s not it. I unpack what the data really says, why companies are at risk even if they pick the right tech, and shine a light on what every individual should take away from it.

⸻

The Model Landscape Is Shifting Fast

The hype around GPT-5 crashed faster than the Hindenburg, especially since hot on the heels of it DeepSeek 3.1 hit the scene with open-source power, local install options, and prices that undercut the competition by an insane order of magnitude. Meanwhile, Liquid AI is rethinking AI architecture entirely, creating models that can run efficiently on mobile devices without draining resources. I break down what these shifts mean for businesses, why cost and accessibility matter, and how leaders should think about the expanding AI ecosystem.

⸻

Google Mango: A Breakthrough in Complexity

Google’s has a new, also not so new, programming language, Mango, which promises to unify access across fragmented databases. Think of it as a universal interpreter that can make sense of siloed systems as if they were one. For organizations, this has the potential to change the game by helping both people and AI work more effectively across complexity. However, despite what some headlines say, it’s not the end of human work. I share why context still matters, what risks leaders need to watch for, and how to avoid overhyping this development.

⸻

A Positive Use Case: Sales Ops Transformation

To close things out, I made some time to share how a failed AI initiative in sales operations was turned around by focusing on context, people, and process. Instead of falling into the 95%, the team got real efficiency gains once the basics were in place. It’s proof that specialized AI can succeed when done right.

⸻

If this episode was helpful, would you share it with someone? Leave a rating, drop a comment with your thoughts, and follow for future updates that help you lead with clarity in the AI age. And, if you’d take me out for a coffee to say thanks, you can do that here: https://www.buymeacoffee.com/christopherlind

—

Show Notes:

In this Weekly Update, Christopher Lind breaks down MIT’s claim that 95% of AI pilots fail, highlights the major shifts happening in the model landscape with DeepSeek and Liquid AI, and explains why Google Mango could be one of the most important tools for managing complexity in the enterprise. He also shares a real-world example of a sales ops project that proves specialized AI can succeed with the right approach.

Timestamps:

00:00 – Introduction and Welcome

01:28 – Overview of Today’s Topics

03:05 – MIT’s Report on AI Pilot Failures

23:39 – The New Model Landscape: DeepSeek and Liquid AI

40:14 – Google Mango and Why It Matters

47:48 – Positive AI Use Case in Sales Ops

53:25 – Final Thoughts

#AItransformation #FutureOfWork #DigitalLeadership #AIrisks #HumanCenteredAI