Generative AI has become the big bet of the year for many companies. But new research shows that enthusiasm often ends up in disappointment. MIT's study out of 300 AI implementations reveals that only 5% of generative AI projects manage to increase revenue quickly. McKinsey Global Survey confirms the picture — 78% of organizations use AI, but over 80% notice no difference in the company's performance. Goldman Sachs question whether the huge investments will ever pay off.
What does this say to you as a leader? That AI is useless?
No - it means that success is about much more than technology.
MIT researcher Aditya Challapally who led the study explains why:
"Generic tools like ChatGPT work well for individuals because they are flexible. But in companies, they remain ineffective because they don't learn from or adapt to workflows."
You will recognize this if you have worked with other change processes. Technology is rarely the bottleneck. It is the organization's ability to embrace change that determines success.
The Goldman Sachs report shows that despite the tech giants' planned investments of over $1000 billion in AI infrastructure in the coming years, there is significant skepticism about whether current AI ventures are delivering results that justify the costs.
The report presents mixed views - while some analysts are optimistic about AI's long-term potential, skeptics point to the fact that the technology cannot yet solve complex enough problems to justify the huge costs. As an example, Goldman Sachs mentions that in its own internal tests they found that AI can update historical data faster than manual work, but at six times the cost.
The report lays bare an industry in conflict with itself. Optimists see potential for 25% automation of all existing tasks with a 9% increase in productivity, while skeptics warn that only 4.6% of tasks will actually be cost-effective to automate within ten years. Goldman Sachs's own research shows that AI adoption "remains modest" across most industries, and that basic tasks such as text summarization often yield "incoherent and meaningless results."
While Nvidia and other infrastructure companies are reaping record profits from the AI buildout, most other companies are struggling with proving ROI on their AI investments. The question remains: will this massive investment ever pay off, or are we witnessing the build-up of a new technology bubble?
The MIT study reveals something interesting: over half of AI budgets go to sales and marketing tools. But the biggest payoff comes from automating administrative processes - eliminating manual labor, reducing outside consultants and streamlining operations.
It reminds me how many approach quality work. They focus on documentation instead of improving processes because that's what they think they should do. Goldman Sachs's Jim Covello, head of Global Equity Research, points to the same problem:
"AI technology is expensive, and to justify the costs, the technology needs to be able to solve complex problems, which it is not built to do."
At the same time, McKinsey's research underlines that AI transformation is often expensive and resource-intensive. This explains why so many AI projects get stuck - without clear prioritization from management, they compete for the same resources as any other initiative.
One reflection I make is that for many it is too early to talk about "AI transformation" when they have not yet managed "digital transformation."
McKinsey research shows that organizations use AI for an average of just 3 business functions - despite the fact that the technology could be used much more widely for better results.
Despite the high failure rate, however, there are clear success factors to learn from. Companies that purchase AI solutions from specialized vendors succeed 67% of the time, while those who build their own solutions succeed only 33% as often.
Challapally explains:
"AI startups led by young entrepreneurs have seen revenue jump from zero to $20 million in a year. That's because they pick a pain point, focus intensely, and collaborate smartly with companies that use their tools."
McKinsey research shows that clear KPIs for AI solutions have the greatest impact on performance and that the CEO's monitoring of AI governance is also a key factor.
Whether you lead a company with 50 or 500 employees, the same principles apply.
As with ISO certification, success with AI is about management commitment, clear processes and systematic follow-up. You don't have to be a technology expert, but you have to be involved in driving the change.
Start small and focused.
Pick a specific - and real - problem where AI can make a difference.
Measure the result carefully.
And remember -- the MIT study's authors are clear:
"Testing AI is easy, but making money from it is hard."
It requires the same discipline as other successful organizational changes.
The research shows that the future belongs to specialized AI tools that solve specific business problems, not generic solutions.
At AmpliFlow, we are currently developing AI capabilities in our management systems and operations management software — but we take our time and meet the real needs of our customers.
Do you have issues within our domain that you think AI could make a difference? Get in touch!
