Despite pouring an estimated $30-40bn into generative AI, 95% of organisations are failing to see any measurable financial impact, according to MIT’s latest report.
The study, The GenAI Divide: State of AI in Business 2025, examined 300 public implementations alongside executive interviews and surveys, and found that most companies piloting the technology had yet to achieve returns, with only 5% extracting millions of dollars in value.
The divide is not explained by regulation or model quality, the report argues, but by approach.
It states that widely adopted tools such as ChatGPT and Microsoft’s Copilot are enhancing individual productivity but have not translated into profit-and-loss gains.
More advanced enterprise-grade systems, whether custom-built or bought from vendors, are faltering in practice.
While 60% of organisations evaluated such tools, only 20% reached the pilot stage, and just 5% moved into full deployment. Most were abandoned due to brittle workflows, limited contextual learning and poor fit with existing operations.
Large firms, despite leading in pilot activity, have been slow to scale up, the report says.
Investment tends to flow into visible, top-line functions rather than high-return back-office processes, while companies working with external partners are achieving higher success rates than those relying solely on in-house efforts. The core obstacle, however, is learning; it continues.
The report claims that systems fail to retain feedback or adapt to context over time, leaving them unable to deliver sustained value.
Those on the winning side of the divide are demanding process-specific customisation and evaluating tools against business outcomes rather than software benchmarks.
Vendors meeting these criteria are already securing multimillion-dollar deployments within months. Some adopters are reporting reduced spending on outsourcing and agencies, while others cite improvements in customer retention and sales conversion.
At the industry level, disruption remains limited. Technology and media stand out as the only sectors showing structural change, with most others confined to pilots in support, content creation and analytics.
Kriti Sharma, chief executive of Nexus Black at enterprise software group IFS, said the findings should be seen in context: “In energy, utilities and manufacturing, AI has already moved beyond flashy pilots into the engine room of business, where the gains are real,” she said, adding that 88% of organisations in these sectors were already reporting profits from industrial AI.
But Sharma warned that progress depends on workforce readiness. “99% of the global workforce will need reskilling to make this shift,” she said.
“The winners will be those who match cutting-edge tech with sharp execution, reskill at speed, and close the AI trust gap.”
