Why Are We Panicking About AI's 95% Failure Rate?

When Lady Gaga was on her seemingly endless press tour for A Star Is Born in 2018, she repeated one line over and over again:

There could be 100 people in the room, and 99 don’t believe in you….but one person does. And it can change everything. 

Gaga’s “hundred people in the room” phrase became a campy internet sensation overnight. However, it was the first thing I thought of when the infamous MIT NANDA report claiming that enterprise AI pilot programs have a 95% failure rate came out.[*]

Well, maybe it was the second thing I thought. The first thing I thought was, Oh, s*%t.

The Truth About THAT AI Success Rate Study

There are a lot of things to question in the study, as Ethan Mollick, the Wharton School professor and author of the book “Co-Intelligence: Living and Working with AI” points out.[*] The most glaring issue is that the study defines “failure” as a lack of meaningful, sustained financial impact within 6 months. Then there’s the fact that the 95% number was the result of “structured” interviews (whatever that means) from just 52 businesses – plus 153 survey responses from attendees of 4 different industry conferences. No company reporting figures, no standardized metrics of success across interviewed organizations.

But what stood out more to me was the willingness to simply gloss over the reality that the 5% of GenAI pilots that do reach deployment have already generated millions in value. Maybe I’m wrong, but that seems like a win to me. Whether you’re implementing a new technology, running experiments in a lab, or even creating an OnlyFans account to make some extra cash, most of us don’t assume instant overnight success when we try something new. We expect a high rate of failure. We expect things to go wrong. We expect initial losses. But given the potential payoff of even just one or two successes? All those failures are well worth it.

I’ll take a high failure rate–and even initial losses–if it brings me closer to finding the one GenAI use case or implementation strategy that makes my company millions and my customers happy.

The Lady Gaga Approach to AI Adoption

Here’s where it gets back to Lady Gaga. In her mind, just one person (her costar Bradley Cooper) believed in her ability to successfully hold her own in a blockbuster Hollywood remake of one of the most famous films of all time. While we can debate the veracity of those claims, she seemed to genuinely believe that 99% of people wanted her to fail spectacularly. Instead, she won an Oscar, Grammy, a BAFTA Award, and a Golden Globe. The film itself received 283 nominations, secured 111 wins, and generated nearly $180 million in studio profits.

So why are so many so willing to completely ignore the impact of a 5% success rate? I find it hard to believe that today’s IT teams really expect even half of their AI pilots to make it to deployment. Most companies are still in the experimentation and testing phase of their AI adoption journey, and I’m willing to bet that very few will reach full-scale AI implementation even by the end of 2026. Many only started taking GenAI adoption seriously at the start of 2025, when it became clear that a resistance to using AI for CX would result in far more harm than good. Instead of lamenting the 95% failure rate, why not celebrate the 5% that actually demonstrate AI’s potential?

Embrace The 5% Impact 

In August, I attended a Nextiva Customer Innovation Day in NYC. One of the attendees expressed frustration at all of the fluff and hype surrounding AI in business communication software. They were tired of the AI noise, the over-the-top promises of “transformation,” “modernization,” and “next-level CX.” They weren’t looking to change everything about their contact center software overnight. Instead, they wanted small, measurable, and realistic impacts that actually moved the needle without completely disrupting operations. They didn’t want a high AI pilot program success rate. They wanted a 5% success rate with an outsized impact.

Lady Gaga didn’t need 100 people in a room to believe in her to win all those awards, and your company doesn’t need every AI initiative to succeed to improve the customer experience and generate profits. Instead, you need to find your “one person in the room”: that single use case, perfect implementation strategy, or breakthrough AI application that changes everything.  Stop making it about the 99 AI initiatives that don’t work out, and focus on finding the one that does.