The dust has settled from the recent report out of MIT which featured a study that contained a whiplash-inducing headline: 95% of AI Initiatives are Failing with a capital F.
The market briefly felt an impact. The study is revealing, and there are lessons to glean, but for most of us bullish on AI, we are going to be just fine if we take a closer look.
I took some time to carefully review the study and I think it has much guidance to offer. I practice what I preach at Learnit—critical thinking skills will be incredibly important in the age of AI. We’re going to see more bombshell studies, predictions, PR stunts, as well as better and faster models as the months and years go by. I urge you to not lose your head when each of these fresh new pieces come out. I urge you to read the studies and not just the headlines or hot takes. There are a lot of people invested in making you lose your head. Don’t let them win.
Here are five takeaways from that study that business leaders can use today.
90% of employees surveyed said they use AI for daily work. Only 40% of companies surveyed said they have officially rolled out subscription-based LLM tools. This means that employees are using AI whether you like it or not, and (worse) whether you allow it or not.
The risk here is that employees might be using sensitive information in unsecure systems. They are doing this because they are finding that the tools are making them work more efficiently.
Organizations that are slow to meet employees where they are at are most at risk of letting the Shadow AI phenomenon create problems.
If your organization has not adopted a philosophy or legal boundaries around what employees can do and can’t do, the time to implement communication on this is yesterday.
What’s better—at the individual level ($20/month) the best and fastest gains were made for efficiency well over the institutional tools that are built for many, many times that rate.
The case is clear here: Employees are using AI. Train them to ensure they use it responsibly and in a way that aligns with your organization’s goals and values.
This one is going to be brief.
While some internal AI builds can succeed, the study found that two thirds of internal builds failed to reach deployment.
The companies who made investments through partnerships succeeded at a 67% rate. As savvy as you might be internally, consider looking for a great partner in creating your AI solutions.
There were some marginal gains with AI deployment in Front Office such as slightly better Customer Retention (10%) and some better lead generation (40%), but "back-office deployments often delivered faster payback periods and clearer cost reductions."
To be clear: the best organizations found a balance between front and back office, but back office BPO elimination is where the top 5% of organizations found millions in savings. “Back-office automation may offer more dramatic and sustainable returns for organizations.”
This looks like employing systems that can help
It can be tempting to look towards the front office first to flash your new AI Initiative, but do not miss the opportunities in the back office.
Here’s the one to remember when you are making your AI Initiative: “The core barrier to scaling is not infrastructure, regulation, or talent. It is learning. Most GenAI systems do not retain feedback, adapt to context, or improve over time."
If your AI System cannot remember or cannot adapt and improve with the changing environment then you have a static tool that will not perform any better than your $20/month subscription. This is borne out in the data. Even when employees have an institutionally built tool ingrained into the workflow, they still prefer using external tools like ChatGPT.
"The most forward-thinking organizations are already experimenting with agentic systems that can learn, remember, and act autonomously within defined parameters."
Because so many of the deployed tools are not agentically learning employees are using them only for short, low-context tasks. When given the option for high-value tasks, they preferred a junior associate performing the task over AI.
If you’re making an investment, make sure it can learn and adapt dynamically.
The biggest wins came from organizations that targeted a specific pain point around a process and leveraged a learning capable system.
Examples of successful categories for narrow applications include:
The companies that are doing the best are looking for small, narrow victories before expanding. This means taking a very specific approach in what you are deploying AI to do for your organization.
“Tools with low setup burden and fast time-to-value outperform heavy enterprise builds.”
Even Sam Altman mentioned that we might be in an AI Bubble. But after the dot com bubble, the internet became better. After 2008, houses were still here and they are still a great investment. No matter what happens with AI, the technology is going to be a part of work.
Don’t lose your head.
The 95% failure rate is a wake-up call for organizations trying to boil the ocean instead of heating the kettle.
The 5% of companies succeeding are getting strategic. They're meeting employees where they already are with AI, partnering instead of building from scratch, automating their back offices before glamorizing their front offices, investing in systems that actually learn, and proving value in small wins before betting the farm.
The employees using shadow AI aren't waiting for your permission. The question is whether you'll lead the change or let it happen to you.
Want our articles in your inbox? Sign up for our blog newsletter to never miss out!
