Is your AI project doomed to fail?

A number often floated in IT circles is that 70% of projects fail. In fact, it led to a discussion this week on LinkedIn that got me thinking about the figure. I asked Google search, Claude, Gemini, ChatGPT and Perplexity if the number was valid, and if there was research to back it up. 

They all said that there was some truth to the number from various reputable research firms, but the level of variability around what you mean by failure, and the types of project, make it challenging to nail down just how many projects fail. Even defining exactly what you mean by “failing” is hard – did it never get completed, run out of money, lose its primary sponsor, get dumped for something more modern? 

Early in my career I would have been shocked hearing a number like this being thrown around. But after 25 years covering the enterprise, it's become easier to accept. No matter how important the technology seems to be, I've learned that implementation is challenging, even with the best of intentions.

That’s not to say that there aren’t companies that have successfully transformed, grown and managed to dig out of technical debt to get to the promised land, but it’s not easy and the people involved must often push through multiple obstacles before achieving success. 

As CXOTalk podcast host Michael Krigsman told me in a 2016 TechCrunch article on digital transformation, it takes real leadership: “The successful executives are able to embrace change. This is a very key point and it’s really the most difficult thing about this,” he told me at the time.

His quote is still surprisingly relevant, and perhaps even more so when you look at all the money being thrown at generative AI projects at the moment. Companies are spending huge sums, but as I wrote in an article a couple of weeks ago, the current research suggests that organizations are struggling to get out of the proof of concept stage and into widespread distribution.

"The successful executives are able to embrace change. This is a very key point and it’s really the most difficult thing about this."

One of the research reports published by Deloitte in January found that “over two-thirds of respondents [to a recent survey] said that 30% or fewer of their current experiments will be fully scaled in the next three to six months.” That means that 70% won't make it. That's a familiar looking number.

It’s clear that IT projects tend to fail at a high rate, regardless of the exact percentage or how you define failure. AI projects are exponentially harder, both from a technical perspective, and because there is a kind of FOMO madness around them. That puts even more pressure on a complex project to succeed, while creating an arms race that is impossible to win.

Large organizations move slowly, and the world is moving quickly, which only adds to the challenge for IT departments to keep up, making it even more difficult to implement these projects successfully. But the good news is that there are visionary leaders who have moved their organizations forward. Look at the 30% of projects in Deloitte survey that are successfully moving into production as your guide. While it’s clearly not going to be easy, it's not impossible either, and you can learn from resourceful leaders who are getting it done.

Photo by Jordan Harrison on Unsplash