AI Industry 07 Jun 2025
AI Project Failures - Whitepaper Discussion
Here's a good one I've seen making the rounds: From a study done by Rand - "By some estimates, more than 80 percent of AI projects fail".
- Chris
Making the rounds - AI Project Failures
This is twice the already-high rate of failure in corporate information technology (IT) projects that do not involve AI. There are numerous reasons for these failure to launch or sustain initiatives, but the theme really revolves around a simple quote: "AI is not a magic wand that can make any challenging problem disap- pear; in some cases"
The reality? Organizations are too excited about the potential, the idea of AI, to stop and consider the investment required to take advantage of AI. Probabilistic algorithms require clean and accurate information and clear and concise goals that the algorithms are set to solve. Too many organizations are buying into a product that they too little understand, and inevitably waste the investment.
Nearly all businesses have troves of data consisting of a mixture of databases, spreadsheets, images, pdfs and in many cases paper. Each and every piece is required to make a full picture of your data, your customers, your product. If any piece is missing or damaged, it can be impossible for a human, much less an AI to make any determination on the state of your business. But this is the state most industries start their AI aspirations from.
Of course, data is inconsequential without defined goals and metrics. The famous answer to the meaning of life, 42, is a great goal, but as the story goes on, the real meaning is in the question. What does your company need to monitor to achieve its goals. What are the goals ? Can you define success? If you can't, then how is a computer?
Gen AI is an addictive substance. Write a few words and you can get a funny image or a video. But that usually does not directly translate into a successful campaign, and at an obscene cost in compute time. For anything real, meaningful, you need to power it with more insight and more data, and if you lack those assets, your grand AI plans are likely to fail.
As always, even with advanced thinking computers, the real story is still -- garbage in, garbage out.