In brief: Three out of four large enterprises use AI in production, but half cannot consistently demonstrate return on investment because clear target metrics and governance structures are missing.
A study by Plug and Play shows that while AI in large enterprises has made the leap into productive use, verifiable return on investment remains unmeasurable for half of respondents. For CEOs, this means that investment decisions are often made without reliable performance tracking.
According to the study, 74 percent of the large enterprises surveyed already use AI in production, yet half say they cannot consistently measure return on investment. Amit Patel, Partner at Plug and Play, sums it up as follows: “Enterprise AI has crossed the threshold into production, but not the threshold into demonstrable value.” The study cites missing baseline values prior to deployment, inconsistent calculation methods, measurement periods that are too short, and persistent underinvestment in the underlying data foundation as causes. The gap is particularly pronounced among companies that so far only deploy AI within a single business unit: 74 percent of them state that value assessment is either premature or not tracked at all. The survey draws predominantly on Fortune 500 and Forbes Global 2000 companies, and therefore primarily reflects the top end of the market. Only 5 percent of respondents describe themselves as fully AI-native, 37 percent deploy AI within a single business unit, and a further 32 percent deploy it across multiple units.
Priorities in vendor selection have shifted markedly: 92 percent cite data protection, explainability and compliance as the most important selection criterion, ahead of pure performance (74 percent) and flexibility (53 percent). Organizational responsibility for AI is also inconsistent: only 21 percent of companies place responsibility with a central AI leadership role or dedicated center of excellence, while 37 percent leave it to individual business units. For corporate leadership, this means that without a clear governance structure, control over AI investments remains fragmented, and success metrics are gathered in a decentralized, inconsistent manner.
Industry publication MarketScale places the findings within two parallel market developments. On one hand, according to Forbes author Gaurav Sharma, industrial conglomerates such as Schneider Electric and Siemens are consolidating the industrial AI market through several multi-billion-dollar acquisitions, meaning AI capabilities are increasingly being embedded directly into infrastructure platforms already in use rather than procured separately. On the other hand, a further Forbes analysis by Jason Snyder warns that companies are increasingly outsourcing core business knowledge into models and platforms they do not own themselves, without first carefully reviewing contractual terms on data sovereignty, model updates and exit rights.
MarketScale specifically recommends that IT and procurement leaders scrutinize existing contracts with AI vendors just as closely as ERP or cloud contracts, examining data portability, audit rights, and the handling of proprietary training data in the event of termination. In addition, every production AI deployment should be tied from the outset to a specifically named business metric, rather than measuring effectiveness only after the fact — a point that, given the study’s findings, has direct relevance for investment governance at the CEO level.
Source: www.it-daily.net · Published August 17, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.8.3.