Study warns of ‘learning gap’ in enterprise adoption as most projects stall at pilot stage
A new study from the Massachusetts Institute of Technology has revealed that the vast majority of corporate generative AI pilots are failing to generate meaningful financial returns, despite widespread investment.
The report, The GenAI Divide: State of AI in Business 2025, published by MIT’s NADA initiative, found that 95% of pilots stall at early stages and never progress to scaled adoption. Only 5% of projects achieved rapid revenue growth.
Based on 150 interviews with business leaders, a survey of 350 employees and an analysis of 300 public AI deployments, the study highlights a growing divide between successful AI integrations and those that remain stuck in experimental mode.
AI pilots stalling at scale-up stage
Aditya Challapally, lead author of the report and head of the Connected AI group at MIT Media Lab, said failures were less about the quality of the underlying models and more about how organisations attempt to use them. He described the issue as a “learning gap” between the tools and enterprise workflows.
“Unlike consumer tools such as ChatGPT, which adapt flexibly to individual use, enterprise AI requires careful integration into existing systems. Without that, pilots rarely scale,” Challapally explained.
One key finding is that more than half of corporate AI budgets are currently directed at sales and marketing use cases, despite the strongest returns being reported in back-office functions such as business process automation, reduced outsourcing and operational efficiency. MIT researchers suggest that this mismatch reflects a lack of strategic clarity in many organisations’ AI agendas.
Success rates also vary depending on how companies source their technology. AI tools acquired through external suppliers or partnerships succeed around two-thirds of the time, while internally developed systems succeed only one-third of the time. This raises questions for industries such as financial services, where many firms are pursuing proprietary in-house models.
Workplace shifts without mass redundancies
The report also points to emerging workplace changes. Rather than mass redundancies, firms are increasingly opting not to replace departing staff in customer support and administrative roles, many of which had previously been outsourced.
Another trend highlighted is the rise of “shadow AI” tools such as ChatGPT adopted by employees without company sanction, which makes it difficult for leaders to track the technology’s real impact on productivity or profit. At the same time, some organisations are beginning to experiment with agentic AI systems that can act autonomously within set boundaries, suggesting the next stage of enterprise adoption.
MIT’s report concludes with a call for line managers, not just central AI teams, to take a lead in integrating AI into daily operations. It also recommends that firms prioritise tools that can evolve with organisational needs, rather than static pilots.