95% of generative AI pilots at companies are failing

A new MIT study finds that 95% of organizations aren’t seeing any financial return from their generative AI investments.

Only a small fraction—about 5% of companies running integrated AI pilots—are reporting meaningful gains, with those few reaping millions in value. For the rest, AI adoption has so far produced little to no impact on profits.

While more than 80% of businesses have tested or piloted tools like OpenAI’s ChatGPT or Microsoft’s Copilot, and nearly 40% say they’ve rolled them out more broadly, most applications are focused on boosting individual efficiency rather than moving the bottom line.

The gap comes down to design flaws and limitations: workflows that break easily, systems that don’t adapt to context, and weak alignment with daily business operations. As the researchers note, today’s generative AI tools still can’t retain user feedback, adjust to real-world nuance, or continuously improve like humans can.

The report also challenges the assumption that AI will soon trigger mass layoffs. Instead, it argues that near-term impact is more likely to come from external cost-cutting measures—such as optimizing processes or reducing third-party spending—rather than wholesale restructuring of internal workforces.