Three independent research groups measured how these tools perform inside real businesses in 2025. The findings converge: most projects produce nothing measurable, and the small fraction that works retains context instead of starting from zero every session. That second part is the reason Third Vector exists.
of companies abandoned most of their AI initiatives in 2025, up from 17% the year before.
S&P Global Market Intelligence · Voice of the Enterprise: AI and Machine Learning, Use Cases 2025 · March 2025
A survey of just over a thousand organizations across North America and Europe. The same study found the average organization scrapped 46 percent of its proof-of-concept projects before they reached production. Starting pilots is not the hard part. Getting anything durable out of them is.
of enterprise pilot programs produced no measurable return.
MIT Media Lab, Project NANDA · The GenAI Divide: State of AI in Business 2025 · August 2025
A preliminary report built on 150 executive interviews, a survey of 350 employees, and 300 public deployments. The headline number has been debated and the methodology has limits, so read it as one data point among several rather than settled fact. What has held up better than the figure itself is the report's explanation for the divide: the tools that failed did not retain feedback or accumulate context, and the small group that succeeded used systems that learn.
of organizations qualify as high performers, attributing more than five percent of profit to these tools.
McKinsey & Company · The State of AI in 2025: Agents, Innovation, and Transformation · November 2025
A survey of 1,993 respondents across 105 countries. Only 39 percent of organizations reported any enterprise-level earnings impact at all. The gap between using the tools and getting paid for them is the whole game.