Generative AI is causing excitement but not success for most enterprises. This needs to change quickly, but it will take some work that enterprises may not be willing to do.
The public cloud market is seeing exploding growth, and it’s easy to understand why. The interest in generative AI has sent enterprises running to their public cloud console to allocate even more resources, including data storage and compute, which tend to be higher-end and more costly.
You don’t need to look far for the disappointing stats. Gartner estimates that 85% of AI implementations fall short of expectations or aren’t completed. I see the same thing in my practice: Projects start and then stop, many never to be resurrected. You can google all the other reports of AI bad news; the general trend is that companies are good at spending money but bad at building and deploying AI.
Seen this movie before?
Reports indicate a significant shift in how cloud technologies are deployed, led by demand for generative AI with its intensive computational requirements. The increased reliance on cloud services to host, train, and deploy AI models illustrates the symbiotic relationship between AI innovations and cloud infrastructure. Organizations have invested heavily in cloud-based solutions to accommodate the complex requirements of advanced AI models, pushing the limits of cloud capacity and capabilities.
Unfortunately, AI is failing everywhere. The abandonment rate of projects reflects a broader trend of resource misalignment and strategic oversights. The rapid advancements in AI capabilities have been matched by increased complexity and specificity of data requirements. Many organizations need help sourcing and managing high-quality data for successful AI deployments, which has become an obstacle that most enterprises must overcome.
Data is the problem
Poor data quality is a central factor contributing to project failures. As companies venture into more complex AI applications, the demand for tailored, high-quality data sets has exposed deficiencies in existing enterprise data. Although most enterprises understood that their data could have been better, they haven’t known how bad. For years, enterprises have been kicking the data can down the road, unwilling to fix it, while technical debt gathered.
AI requires excellent, accurate data that many enterprises don’t have—at least, not without putting in a great deal of work. This is why many enterprises are giving up on generative AI. The data problems are too expensive to fix, and many CIOs who know what’s good for their careers don’t want to take it on. The intricacies in labeling, cleaning, and updating data to maintain its relevance for training models have become increasingly challenging, underscoring another layer of complexity that organizations must navigate.
Usually, data issues result from past mistakes made by predecessors, such as pushing many of the processes and key data elements to ERP systems or chasing hype-driven trends, such as data warehouses. As a CIO friend told me recently, “I’m not going to take a hit for somebody else’s bad decision.”
The cloud won’t save you
Despite these challenges, integrating AI with cloud computing remains a key focus area, providing essential infrastructure for scaling AI initiatives. Companies continue to explore cloud solutions to support their AI ambitions. However, we know now that the return on investment has been slower than expected.
The disparity between the potential and practicality of generative AI projects is leading to cautious optimism and reevaluations of AI strategies. This pushes organizations to carefully assess the foundational elements necessary for AI success, including robust data governance and strategic planning—all things that enterprises are considering too expensive and too risky to deploy just to make AI work.
The understanding here is that the cloud won’t save you. This is not an issue with the platform; this is an issue with the knowledge of data assets and resources needed to make generative AI work for enterprises.
I suspect that this will lead to haves and have nots in the world of AI. Those who can get their data in good order and effectively use AI can use generative AI as a strategic differentiator that will take the company to the next level. Others will watch and fall by the wayside.
Cloud providers will grow over the next few years, much like we’re seeing now. However, unless they can teach their customers how to define an AI strategy that can overcome the many failures, their market will contract again. At least we’ll know why.
The reasons enterprises suck at generative AI and tank their projects are well understood. This is not a mistake that analysts and CTOs can’t explain. We know why AI projects are taking a dirt nap, and enterprises don’t seem to be willing or able to invest in a fix. I suspect they will have to, sooner or later, and I hope some CIOs have the political courage to address things head-on, cloud or no cloud. That’s the only way this still works.
