Report: 80% of AI projects fail, doubling regular IT project failure rates

Facepalm: We're told that it's vital any firm in the tech industry, and plenty of those that aren't, embrace AI, lest they be left behind and potentially go out of business. Yet according to a new report, a stunning 80% of all AI projects fail, twice the failure rate of information technology projects that do not involve artificial intelligence.

Companies are investing billions of dollars into AI and machine learning, despite the slow returns and plenty of failures. The RAND Corporation wanted to find out what was behind this high failure rate of 80%, so it interviewed 65 data scientists and engineers with at least five years of experience in building AI/ML models in industry or academia.

The study identified the five leading causes of AI projects failing. The first and most common of these was industry stakeholders often misunderstanding or miscommunicating what problem needs to be solved using AI and what the technology is capable of achieving.

The amount of hype surrounding generative AI means some executives believe its use can magically transform a company for the better. They fail to comprehend how the tech can be applied to their business, what resources are required to implement it, and how long the process will take.

"Here's where we lost all our money after investing in an AI system for a dog grooming salon"

One interviewee said "Often, models are delivered as 50 percent of what they could have been" due to changing priorities and unrealistic timelines.

Another key failure point in AI projects is organizations lacking the necessary data required to adequately train an effective AI model. "80 percent of AI is the dirty work of data engineering," an interviewee said. "You need good people doing the dirty work – otherwise their mistakes poison the algorithms."

There's also the problem of data scientists and engineers focusing on using the latest and best version of AI tech instead of asking if its use would solve any actual problems faced by users.

The other two identified factors were organizations lacking adequate infrastructure to manage their data and deploy completed AI models, and AI being applied to problems that are too difficult for it to solve.

Barring a few exceptions, there have long been questions over the real-world application of some AI projects. It's an important issue to address; one could argue that Microsoft rushed to implement the AI-powered Recall into Windows without thinking about how users would react – the feature came in for huge criticism and was delayed.

There have been plenty of other studies and reports that don't look good for AI businesses. Earlier this month, we heard that just including "AI" in product descriptions makes them less appealing to consumers. A recent poll also showed most people would not pay extra money for hardware with AI capabilities and features. But the worst news for businesses is that reaping the financial rewards from investing in the AI industry is taking longer than expected.