The large majority of generative AI pilot programs deliver little or no measurable impact on company revenues.
That’s according to a new MIT report that examined over 300 public corporate AI implementations, surveyed 350 employees, and talked to 150 industry leaders to find that just 5% of AI pilots earn their companies significant growth.
That’s a dismal success rate for the highly-hyped generative AI technology, and could easily be considered yet another sign that the AI hype bubble is reaching a tipping point.
95% of GenAI Pilot Projects Fail
The data comes from MIT’s NANDA initiative, in a report titled The GenAI Divide: State of AI in Business 2025.
According to coverage from Fortune, the core problem driving failure for 95% of companies is a “learning gap” for tools and organizations. Standard chatbot solutions like ChatGPT work fine for individuals, but can’t adapt to company-wide workflows. They stall out when used for enterprise-level learning.
This just in! View
the top business tech deals for 2025 👨💻
What about the few programs that do succeed? According to Aditya Challapally, the lead author of the report, they keep their scope small:
“Some large companies’ pilots and younger startups are really excelling with generative AI. [Startups led by 19- or 20-year-olds] have seen revenues jump from zero to $20 million in a year. It’s because they pick one pain point, execute well, and partner smartly with companies who use their tools.” -Challapally
AI Resource Allocation Is Part of the Problem
Poor resource allocation is another issue, according to the report.
More than half of generative AI pilots are operating with sales and marketing tools. But that isn’t where AI can be most effective, MIT found.
Instead, the biggest ROIs revealed themselves in behind-the-scenes automation: “Eliminating business process outsourcing, cutting external agency costs, and streamlining operations,” as Fortune summed it up.
Still, the larger problem of an inescapable “learning gap” is damning enough. Even better resource allocation can’t fix an artificial intelligence that can’t learn to be more intelligent.
Is the AI Hype Bubble Bursting?
If you’ve followed tech headlines across the past three or so years, you just might be sick of hearing about AI. The new generative AI technology has driven countless billion-dollar investments and has been embedded in seemingly every enterprise software product you’ve heard of.
Now, the hype bubble might be bursting. AI tools have plenty of uses and benefits, but in 2025, the exponential growth of their functionality certainly seems to have tapered off.
A recent restructure of Meta’s AI team and an underwhelming response to OpenAI’s latest GPT model may be a few signs that the market is cooling on the technology. Another sign is the actual market cooling: The AI-reliant NASDAQ Composite recently posted its biggest decline in weeks, with companies like the data miner Palantir falling 9.4% and chipmaker Arm Holdings dropping 5%.
If AI can’t deliver on all its lofty promises, it may drag the economy down with it.
