The world is abuzz about AI, but a recent report from Massachusetts Institute of Technology (MIT) has just thrown out a staggering number: 95% of enterprise AI projects are failing. Now that’s a figure that warrants a double-take.
The report is the culmination of 153 executive interviews, 350 employee surveys, and an examination of 300 deployments. Only a shocking 5% of these AI projects were found to have actually made it to production, demonstrating any real revenue growth. The other 95% found themselves lost in the abyss of being ‘strategic’.
Two distinct sides emerged from this study. On one side, we have successful startups, some even run by the GenZs, who have gone from nada to $20M in revenues within a year. How do they manage this? Their strategy is simple and efficient; identify a single pain point, go all guns blazing to solve it, smartly collaborating as needed, without overcomplicating the process.
On the flip side, large enterprises are moving at a glacial pace due to their myriad ‘innovation pilots,’ most of which fail to take off past PowerPoint. Amazingly, the point of difference isn’t about model quality; the tools used are generally up to the task. The culprit is the learning gap. Many enterprises haven’t figured out a way to integrate these systems into their admittedly complicated workflows, leading to a frustrating state of stagnation.
Interestingly, most companies are investing the lion’s share of their genAI budgets on sales and marketing tools. However, the real untapped gold mine, as per MIT’s data, lies in the less glamorous, back-office automations. These are the tools that can reduce outsourcing, cut agency costs, and streamline operations. It’s not usually something you’d brag about at dinner parties, but this is where the big ROI lies.
The insistence of corporations to create their own AI tools, particularly in finance and regulated spaces, despite data showing twice the success rate for purchased AI tools, is also a pervasive problem. Despite all the odds, corporate entities keep insisting on reinventing the wheel.
The implications on the job market are subtle yet profound. Contrary to popular belief, the advent of AI isn’t sparking widespread layoffs. Instead, positions in admin and support are simply not being refilled as people depart. Meanwhile, employees are finding ways to incorporate AI such as ChatGPT into their workflows, even if the official line remains that it’s unauthorized.
The AI consciousness debate is yet another interesting facet of this discussion. Researchers from tech giants like OpenAI and Google DeepMind are pondering the notion of “AI Welfare”. That is; if AI becomes advanced enough to gain consciousness, should they not have rights? This is an argument that has deeply polarized the industry, sparking debates and conversation across forums.
While this may seem like science fiction, it’s an important
conversation to have. Regardless of whether AI develops feelings or not, developing a sense of respect towards these systems could greatly influence how we interact with them.
The big takeaway here is not that AI is inherently flawed, but that enterprises need to rethink how they approach it. Simplify processes, focus on real problems, don’t be afraid to invest in ready-made tools, and allow department managers to lead the charge. With these changes at the helm, we can expect to see a significant shift from mere pilot initiatives to actual profit-generating operations.
Remember, we don’t necessarily have to fear AI, rather, the key lies in learning to adapt and work symbiotically with these systems. As consumers continue to integrate AI into their lives, and as large brands recognize the value of incorporating AI into their operations, the future of AI looks promising, buzzing with opportunities just waiting to be tapped.
