Not long ago, a report from MIT made waves: 95% of enterprise AI pilots fail to deliver business impact. Only a small handful (five percent) ever make it to real, measurable results.
It’s the kind of statistic that makes executives nervous. Huge investments, endless pilots, and very little to show for it.
But there’s another story playing out quietly, far away from boardroom dashboards. While official projects struggle, employees are already getting value every single day. They’re using tools like ChatGPT and Copilot to help with emails, reports, coding, analysis, research and even client work. Most of the time, their managers don’t even know it’s happening.
Call it Invisible AI if you like. We think of it as grassroots innovation. And it’s undoubtedly working.
“While 95% of AI pilots fail, employees are already proving success every single day.”
The Hidden Value of Invisible AI
The scale of what’s happening under the surface is striking. An MIT-led survey found that people at more than ninety percent of companies are already experimenting with tools like ChatGPT and Copilot on their own, even though fewer than half of those organizations officially provide access. In Canada, nearly eight out of ten office workers say they use these tools in their jobs, yet only a quarter rely on company-approved systems. What leaders think of as “experimentation” is, in practice, already part of daily work.
And the gains are not abstract. Research from the Federal Reserve Bank of St. Louis showed that employees using generative tools were able to free up about two hours each week. In practice, that meant a five percent cut in workload and a meaningful lift in overall output. For every task where they leaned on these tools, productivity per hour jumped by nearly a third.
“Invisible AI isn’t only a risk story. It’s a productivity story.”
Other studies point in the same direction. Developers who worked with GitHub Copilot finished their coding assignments in nearly half the time. In call centres, support agents using digital copilots managed to handle more conversations while improving the quality of service. The most dramatic changes were seen among newer staff: with a little automated assistance, they quickly performed with the skill and confidence of their more seasoned peers.
Professional services firms are noticing similar results. At companies like EY and Grant Thornton, consultants report saving as much as a full working day every week by automating routine parts of the job. Freed from repetitive tasks, teams are spending more time on strategy, analysis, and client-facing work.
Taken together, these examples reveal something important: invisible AI is already producing real, measurable benefits across industries, long before most pilots succeed.
What Enterprises Are Learning
Some organizations have started to catch up. Johnson & Johnson realized that out of nearly 900 pilots, only a small fraction, around 10 to 15 percent, delivered most of the results. They refocused on those high-value projects in supply chain and R&D and began to see meaningful returns.
At Carlyle Group, leadership didn’t try to control everything. They leaned into culture. By making AI training part of onboarding, they turned curiosity into capability. Today, 90% of their employees use AI tools daily.
PepsiCo took another path, building an “AI control tower” to link demand forecasting, logistics, and data into one system. For them, AI became less about experimentation and more about running the business better.
And AstraZeneca didn’t wait for regulators to tell them how to manage risk. They set up their own ethics-based audits and policies so teams could move fast without cutting corners.
The common thread is simple: success wasn’t about buying more tools or running more pilots. It was about focus, culture, and execution.
“AI success isn’t about launching more pilots. It’s about focus, culture, and execution.”
What It All Comes Down To
Here’s the big takeaway: AI isn’t some project you launch. It’s already here, woven into the way people work. Employees are showing that every day, often without permission, often without recognition.
The companies that win will be the ones that don’t ignore this energy. They’ll find it, shape it, and scale it responsibly, through strategy, leadership, governance, and talent. The same goes for consultants, who now have the chance to move from advising to orchestrating.
The divide isn’t between startups and enterprises, or between pilots and production. It’s between organizations that treat AI as an experiment, and those that weave it into their DNA.
The future belongs to the orchestrators.
“The real divide isn’t between startups and enterprises. It’s between organizations that treat AI as an experiment, and those that weave it into their DNA.”
Sources: MIT GenAI Divide | IBM Canada | St. Louis Fed | GitHub Copilot Study | QJE Customer Support Study |
Comments
Add a comment…
No comments, yet.
Be the first to comm
