GarryKillian/iStock/Getty Images Plus via Getty Images
Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
Although 95% of AI projects fail, research shows that successful initiatives focus on infrastructure.
Top hurdles include poor integration, lack of skill sets, and difficulty building in-house AI solutions.
Businesses that successfully implement AI are 85% more likely to have worked with third-party AI providers.
When it comes to AI, most people fit in one of three camps: They are enthusiastic supporters of AI who believe it will rapidly transform everything, or they are still -- somehow -- oblivious to AI, or they are skeptics who find most AI promises are overblown and unrealistic, and that many AI solutions today are broken and unimpressive.
Also: Gen AI disillusionment looms, according to Gartner's 2025 Hype Cycle report
I tend to sit somewhere between an enthusiast and a skeptic. I am extremely skeptical about most of the futuristic claims that tend to come from generative AI's biggest supporters, and I find the tendency of current AI models to make major (even catastrophic) mistakes a reason to pull back on the use of AI in most critical use cases.
On the other hand, I am extremely excited about AI's potential, especially when it comes to many of the practical use cases we are seeing businesses focus on in our research. I guess you could call me an enthusiastic AI skeptic.
Focus your AI efforts
This is why the recent study by MIT, which found that 95% of AI projects fail, was so interesting to me. My skeptical side exulted with an attitude of, "See, this is what happens when you fall for the hype."However, my enthusiast side saw how this study reinforces many of the trends Aberdeen research is showing for how businesses can successfully deploy AI.
Also: Forget plug-and-play AI: Here's what successful AI projects do differently
First, the study found that where businesses are focusing their AI efforts matters. Over half of organizations are using AI to try to build efficiencies in sales and marketing.
Of course, this is understandable. Much of the AI hype revolves around perfect AI case studies of businesses replacing sales and service staff with AI and building amazing campaigns with AI, so many executives get excited about the potential to reduce costs (and staff) with these AI strategies.
But we are already seeing reversals in trends, with early AI pioneers admitting that they are moving back to mainly human sales teams and stories of failed AI marketing efforts proliferating.
Also: Worried about AI's soaring energy needs? Avoiding chatbots won't help - but 3 things could
Interestingly, the MIT study found higher levels of success in AI initiatives that focused on the infrastructure side of businesses. This goes along with what Aberdeen is seeing, as we've previously seen impressive returns for organizations that focus on practical implementations of AI.
In fact, when our most recent survey looked at businesses that have been successful at AI modernization initiatives, we found that the top areas they were focused on were cybersecurity, predictive analytics, automation, and monitoring -- decidedly unsexy areas but where the strengths of AI are greatest.
Why some businesses fail
The other interesting findings from the MIT AI study related to AI challenges and how businesses are addressing them. The study examined why businesses were failing and listed pain points such as poor integration, lack of skill sets, and the difficulty of building in-house AI solutions.
Also: I retested GPT-5's coding skills using OpenAI's guidance - and now I trust it even less
Again, this matches very well with the data from Aberdeen's research into successful AI implementations. We've found that the top challenges for businesses doing AI are security concerns, lack of AI skills and expertise, and issues integrating AI with existing systems.
Our research also shows that businesses that have successfully implemented AI are 85% more likely than their peers to have worked with third-party AI providers to help them overcome skills and integration issues. This also tracks with the MIT study, which found that working with companies to assist in AI deployment was more likely to lead to success.
As one would expect, there's a lot of hand-wringing and Monday morning quarterbacking about this study's findings, which, with headlines saying 95% failure rate, sounds really bad. But as someone who has been in tech for a while, most new technologies have pretty high failure rates in their early hype-filled years.
Also: 71% of Americans fear that AI will put 'too many people out of work permanently'
The findings are also, in some ways, comforting for many businesses. Basically, they reinforce that smart businesses ignore the hype, take a focused approach to AI designed to solve real problems, and work with experienced partners to overcome hurdles.
It's a practical approach that, while it may not look sexy in a headline, provides a blueprint for successful AI deployments.