New research claims that the vast majority (80%) of AI-based projects fail – double the normal failure rate of non-AI-based technology proposals.
A study conducted by the Rand Corporation Only 14% of organizations were found to feel fully prepared to adopt AI, despite 84% of business leaders reporting that they believe the technology will have a significant impact on their organization.
The main reason for project failure was a lack of understanding and communication between stakeholders and technical staff about the intent and purpose of the project. This means that managers often fail to give teams the time and resources they need; it is critical to ensure that leaders and technical teams have the same goals.
Magpie syndrome
Not having the data needed to sufficiently train their AI model was another problem for new projects: insufficient investment in infrastructure to support data governance and model deployment means AI projects take longer and are not as effective.
This echoes the above. Lenovo Researchwhich revealed concerns about the computing power and data resources needed to train models.
Another challenge facing new projects was the eagerness to use the latest technologies rather than focusing on solving real problems for users. Experimenting with new technologies helps drive development, but too often they are used for the sake of using them, rather than when they are most appropriate. The researchers explain that successful projects do not get distracted by looking for the latest AI advances, but instead focus on the problem at hand.
Finally, and perhaps not surprisingly, the report found a tendency to overestimate the capabilities of AI itself. Although investment has increased 18-fold since 2013, it is not a one-size-fits-all solution for automating all tasks, and the technology still has significant limitations. Understanding the capabilities of models is crucial to success.
With so much massive pressure to use AI across a variety of industries, businesses need to keep in mind that AI is an investment like any other and carries serious risks if not fully understood or managed properly.