E39 - 95% of AI Projects Fail. Here's the Part Nobody Told You.

About this episode

95% of company AI projects showed zero return. Not small. Zero. No measurable impact on the bottom line. That number comes straight out of MIT this past year, and the moment it hit, every headline reached for the same lazy conclusion: AI is overhyped, the technology is not ready, it is all smoke.

It is not. You were sold a lie, and it did not come from the technology. It came from the people selling it. Every AI and LLM company has everything to gain by telling you their tool can do things it flat out cannot, because the bigger the promise, the bigger the check you write. And the biggest promise of all is the one you have heard a hundred times: it can replace your people.

In this episode of The Digital Dilemma, Nick Dreyfus breaks down what the MIT NANDA report (The GenAI Divide: State of AI in Business 2025) actually found, why 95% of enterprise generative AI pilots deliver no measurable impact on the P&L, and what the 5% who are winning understand that everyone else missed. If you are a CEO, CFO, business owner, or executive trying to separate real AI ROI from hype, this one is for you.

What you will learn:

• The real reason most enterprise AI projects fail, and why it starts with a message your team hears as "go train your replacement"
• How fear drives your best people to bury the exact workflows that hold the most value
• What AI is genuinely good at inside a business (flag it and stop) and where it becomes a liability (the moment you ask it to think or make the decision for you)
• The invoice example that makes the line between a task and a judgment call obvious
• Why generic, off-the-shelf AI tools stall, and why AI built with an outside partner succeeded twice as often as in-house builds
• The energy and water problem behind AI data centers, and why the answer is to advance the technology, not stop it
• A practical five-step plan to keep your company out of the 95%

Chapters:
00:01 The 95% number and the lie you were sold
02:20 What the MIT NANDA report actually says
04:45 The fear that makes your team hide the real wins
07:08 How the winners flip the message
09:30 The data center energy and water problem, and diamond cooling
11:54 The one thing AI is actually good at
14:22 Why generic tools fail and partners win twice as often
16:52 The model that works: pair the person with an AI assistant
19:14 Your five-step plan to stay out of the 95%
21:40 The honest version nobody is selling you

Questions this episode answers: Why do most AI projects fail? Is AI overhyped? What is the MIT 95% AI study? How do businesses get real ROI from AI? Build AI in-house or with a partner? Can AI replace employees?

The takeaway is simple. AI is not magic, and it is not the enemy. It is a narrow tool with very specific jobs. Stop asking it to think. Start asking it to flag, and keep your best people pointed at the work that actually moves the business forward.

If you want to walk through where that line sits in your business, what is safe to automate and what absolutely needs a human, that is a conversation we have every single week. Bring your two or three worst tasks and let us help you find those hours back.

Book time with our team: calendly.com/nickdre/30min

Nick Dreyfus
Vice President of Business Development, i-NETT

Your team focuses on growth, our team makes that happen.
Subscribe for more on AI, cybersecurity, and technology strategy for business leaders.

#AI #ArtificialIntelligence #AIforBusiness #MIT #GenerativeAI #Cybersecurity #ManagedIT #AIStrategy #TheDigitalDilemma #iNETT #AIROI