A new MIT study, The State of AI in Business 2025, has made waves with a bold headline: 95% of AI pilots fail. If you’re leading a business right now, that number might make you think twice about investing in AI at all.
But here’s the reality: the study isn’t wrong, but it’s easy to misinterpret. And if you’re making decisions for your business, the worst thing you could do is take that headline at face value.
What the Study Actually Says
- High adoption, low transformation. Enterprises are pouring money into AI pilots, but most stall before reaching production or economic viability.
- Static tools fail. The biggest barrier isn’t regulation or model quality. It’s that most tools don’t learn, remember, or adapt to workflows. They feel impressive in a demo, but they don’t get better with use.
- Shadow AI is thriving. Even where official projects fail, employees are using ChatGPT and similar tools on their own and seeing real productivity gains. In fact, while only ~40% of companies officially bought AI subscriptions, more than 90% of employees reported using AI personally to get work done.
- ROI is hiding in plain sight. Companies often chase “flashy” marketing tools, but the biggest returns show up in less glamorous places:
- Cutting reliance on outside vendors or agencies
- Handling routine tasks in-house instead of outsourcing them
- Automating finance, operations, and customer service processes to save time and reduce errors
So the 95% number reflects enterprise struggles more than the tech itself.
Where the Study Falls Short
- Narrow success definition. “Success” was defined as measurable ROI within 6 months. That excludes slower, but real, wins like compliance improvements or efficiency gains that don’t immediately show up on a balance sheet.
- Enterprise-heavy sample. The study’s interviews were weighted toward large companies. Their challenges — multi-layer approvals, year-long procurement cycles, rigid IT processes — don’t map neatly to smaller, more agile businesses.
- Ignored everyday impact. The study acknowledged shadow AI adoption but didn’t count it as “success,” even though it’s already transforming how employees work day to day.
In other words: “95% fail” is more about enterprise paralysis than AI’s usefulness.
How Your Business Might Misread This
If you only skim the headline, you might walk away thinking:
- “AI is too risky. We should wait.”
- “If the big players can’t make it work, how could we?”
- “We’d need enterprise budgets to see ROI.”
Those are exactly the wrong conclusions.
What to Do Instead
- Start narrow. The winners in the study focused on specific workflows. For your business, that might mean automating client intake forms, streamlining reporting, or creating a content assistant.
- Look beyond marketing. Sales tools get the hype, but automating back-office work, such as finance, contracts, and operations often deliver faster payback.
- Choose systems that learn. The companies crossing the “GenAI Divide” don’t just use tools that generate. They choose systems that:
- Remember corrections and preferences
- Adapt to workflows instead of resetting each time
- Integrate with existing tools so adoption is natural
These don’t always require advanced “agentic AI.” Sometimes it’s a custom GPT built around your data, or an integration that connects AI into your CRM or CMS. The point is: the system improves as you use it.
- Trust matters. The study highlighted how enterprises stick with vendors they trust. The same is true for your business: choose a partner who understands your operations, not just a vendor with a flashy demo.
- Use shadow AI as a guide. If your team is already experimenting with ChatGPT or Claude on their own, don’t ignore it. Learn from what they’re doing and look for safe, scalable ways to make those wins official.
The Bottom Line
The MIT report is a helpful reminder that AI isn’t magic, and poorly designed pilots will fail. But the real message is this: don’t confuse enterprise paralysis with AI’s potential for your business.
You don’t need a million-dollar pilot. You don’t need a 12-month rollout. What you need is focus, trusted guidance, and a willingness to start small. That’s how businesses like yours cross the so-called “GenAI Divide.”
At MantyWeb, we help organizations use AI strategically, not as a gimmick, but as a tool to work smarter and more sustainably. If you’re wondering where AI could actually make a difference in your business, let’s talk.