An Industry Full of Failures: Research Shows That Over 80 Percent of Artificial Intelligence Projects Fail

Generative AI has captivated the business world, with large language models like OpenAI’s ChatGPT currently boasting over 200 million weekly active users. AI is becoming an increasingly significant component of IT budgets, with the global artificial intelligence market currently valued at nearly 235 billion dollars, and projections indicating a rise to over 631 billion dollars by 2028. However, IBM’s findings reveal that few artificial intelligence projects deliver the financial value that shareholders expect. In fact, the average ROI is only 5.9 percent—significantly below the usual 10 percent cost of capital.

Most Fail

Additionally, according to research from RAND Corporation, over 80 percent of these AI projects will not succeed—double the failure rate for non-AI-related startups. The global policy think tank spoke with 65 data scientists and engineers who have worked in the artificial intelligence sector in recent years and identified several causes leading to this staggering failure rate.

According to the research, the primary reason for the failure of artificial intelligence projects is the misalignment of goals between key stakeholders. Leadership in AI companies often has a view of what AI can and should achieve that is not grounded in reality. Instead, it is driven by a preconceived notion of what AI is, often fueled, as researchers claim, by Hollywood. This lack of understanding between business leaders and those on the ground means that projects often lack the resources and time needed to achieve their goals.

However, the engineers working on the other end of artificial intelligence are not flawless either. Interviews revealed that data scientists are sometimes distracted by the latest advancements in artificial intelligence and implement them in their projects without considering the value they will bring. This ‘shiny object syndrome’ means that scientists and engineers want to use these new technologies simply because they are the latest. While it is important to stay current with artificial intelligence, teams also need to consider whether this new technology will truly solve the problems they face in their research or merely complicate and entangle them further. FOMO seems to be strong in the AI industry as well.

The research also noted several other reasons, including a lack of properly prepared datasets, inadequate infrastructure, and incompatibility of artificial intelligence with the problem at hand. The study confirmed that these issues are not limited to the private sector: even the academic community struggles with artificial intelligence projects, where many focus simply on publishing AI research rather than looking for real-world applications for their results, researchers claim.

A Lot of Failures

This research is evidence of the many consolidations and failures we see in the artificial intelligence industry. In fact, Baidu’s CEO Robin Li Yanhong stated that China has too many large language models and is wasting a significant amount of resources as they often have little, if any, practical applications in the real world. We can also see this with the number of generative AI patents that China has filed in the past decade, outpacing the U.S. by a ratio of 6 to 1. Yet, despite this, only one Chinese organization, the Chinese Academy of Sciences, made it into the top 20 entities with the highest number of citations between 2010 and 2023.

The research also shows that in the race to advance in AI, many companies rush into building their projects. While they (and their investors) are the only ones bearing the risk of any failed project, it would still be wise for them to carefully consider the failures of other AI projects and the reasons behind them. After all, if artificial intelligence projects fail to deliver on their promises over a long period, the entire industry could collapse and burst like a multi-billion dollar bubble.