You know that warm fuzzy feeling you get thinking AI will solve all your business problems and let you retire early? Yeah, this episode is the cold shower you didn’t know you needed. We’re talking about why most AI projects crash harder than a Segway on launch day, how businesses keep falling into the same traps, and why treating AI like a superpowered intern (instead of your replacement) might be the smarter move.
In this episode:
Why 95% of AI Dreams End in a Cold Shower – Ep 540
Today’s Episode is brought to you by:
Kardon
and
HIPAA for MSPs with Security First IT
Subscribe on Apple Podcast. Share us on Social Media. Rate us wherever you find the opportunity.
Great idea! Share Help Me With HIPAA with one person this week!
Learn about offerings from the Kardon Club
and HIPAA for MSPs!
Thanks to our donors. We appreciate your support!
If you would like to donate to the cause you can do that at HelpMeWithHIPAA.com
Like us and leave a review on our Facebook page: www.Facebook.com/HelpMeWithHIPAA
When you see a couple of numbers on the left side of the text below click that and go directly to that part of the audio. Get the best of both worlds from the show notes to the audio and back!
Why 95% of AI Dreams End in a Cold Shower
[00:53]Most companies dive right into AI, expecting an easy win—like settling into a warm hot tub. But guess what? An MIT study found that a whopping 95% of enterprise AI projects crash and burn, leaving you with a metaphorical “cold shower” instead of that relaxing soak. This whole idea came up when Donna was reading a Substack post by Josh Anderson called “I Went All-In on AI. The MIT Study Is Right,” which talked about those dismal MIT stats on AI failures. This week we discuss, “Why 95% of AI Dreams End in a Cold Shower,” by digging into this messy reality, mixing the study’s findings with what we’re seeing in the real world.
This quote was the most telling:
My own product, built under my direction, and I’d lost confidence in my ability to modify it.
We are seeing the same thing he mentions in the article across our clients and discussions with others. AI is going to be awesome. Everyone is singing that song from the Lego movie, everything is awesome! Management sees dollar signs in promised savings. Then, it turns out it isn’t awesome and everyone just stands around staring at each other with no idea what to do.
The article is great and you should read it. The bottom line he came to in his experiment was the same things we have been saying. It is a very good assistant but you don’t turn over control to them. He used the point that you use AI for augmentation of your work not to abdicate your work.
The GenAI Divide: State of AI in Business 2025 report by MIT says more than just that 5% stat plus it reflects a lot of what we are seeing. Here is the opening paragraph of the summary page:
Despite $30–40 billion in enterprise investment into GenAI, this report uncovers a surprising result in that 95% of organizations are getting zero return. The outcomes are so starkly divided across both buyers (enterprises, mid-market, SMBs) and builders (startups, vendors, consultancies) that we call it the GenAI Divide. Just 5% of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable P&L impact. This divide does not seem to be driven by model quality or regulation, but seems to be determined by approach.
[17:54]Here were the 4 patterns MIT noticed which aligned with the article Donna started out reading. When you go all in without a real plan you will likely fail or at least stall in testing mode.
- Real disruption is limited: Out of the eight major industry sectors, only two are seeing any real structural change from generative AI so far.
- The enterprise paradox: Large organizations are running the most AI pilots, but they’re also the worst at turning those pilots into something that actually scales.
- Budget bias: Companies keep pouring money into flashy, customer-facing AI projects, while the high-ROI, back-office opportunities get ignored.
- Partnership advantage: Teams that work with outside partners or vendors see about twice the success rate compared to companies trying to build everything in-house.
The report also highlighted 5 myths that align with the concepts in the article.
Five Myths About GenAI in the Enterprise
1. “AI will replace most jobs in the next few years.”
Research shows layoffs tied to GenAI are limited and mostly concentrated in sectors already heavily automated. Executives don’t even agree on whether hiring will go up or down over the next 3–5 years.
2. “Generative AI is transforming business.”
Adoption is widespread, but true transformation is rare. Only about 5% of enterprises have integrated AI into workflows at scale, and 7 out of 9 major sectors show little to no structural change.
3. “Enterprises are slow to adopt new technology.”
Not this time. Companies are extremely eager to adopt AI — roughly 90% have seriously explored buying an AI solution.
4. “AI isn’t scaling because of model quality, legal issues, or data risk.”
What’s actually slowing things down is that most AI tools don’t learn over time and don’t integrate well into existing workflows.
5. “Building AI internally is the best path to success.”
The data says otherwise: organizations that partner with external vendors or experts see roughly twice the success rate compared to companies trying to build everything in-house.
Back to our original article from Josh Anderson here. The point that using AI risks us losing what creates the judgment and wisdom that only comes from the scars of experience. You can’t prompt your way to understanding the little things you pick up like the sounds of someone typing the incorrect command on the keyboard over the phone. Or, why that new computer kept shutting down due to a loose card that didn’t look loose.
The Agentic AI Explosion: A HIPAA Nightmare in the Making
[29:32]Employees are using this stuff and that use is untrained, unvetted and definitely not properly documented and secured. Shadow AI is going to be a bigger problem than Shadow IT ever has been. We know this isn’t going to get better as things like the Agentic AI marketplaces start offering wonderful solutions!
David just found one we seriously doubt has been vetted and probably doesn’t even understand that being a BA is anything more than signing a BAA. But wait, that opens up a whole new door.
[36:12]We recently discussed the new things coming out from HSCC on AI in 2026 and we have even more coming at us since then. Look for us to do several more episodes on topics about implementing AI safely and securely in the new year. We don’t have any options just like you guys. We have to deal with it with governance and all that comes with that.
Use these 3 MIT guides when implementing AI in your organization
Researchers Uncover 30+ Flaws in AI Coding Tools Enabling Data Theft and RCE Attacks
HHS Artificial Intelligence AI Strategy
AI may be the future, but right now, it’s a bit like giving your toddler a chainsaw — sure, it can do amazing things, but someone’s definitely going to lose a finger if you’re not careful. So as we barrel into another year of shiny tools, bold promises, and disappearing platforms, just remember: success isn’t about jumping in headfirst — it’s about knowing where the shallow end is.
Remember to follow us and share us on your favorite social media site. Rate us on your podcasting apps, we need your help to keep spreading the word. As always, send in your questions and ideas!
HIPAA is not about compliance,
it’s about patient care.TM
Special thanks to our sponsors Security First IT and Kardon.
