Most enterprises experiment, few scale
A McKinsey note reported on social says nearly two‑thirds of enterprises have experimented with AI agents but fewer than 10% have successfully scaled them, highlighting data foundations as the main obstacle. The statistic was circulated as a signal about the gap between prototypes and production rollouts. (x.com/i/status/2042679160846368989)
Most large companies have tried artificial intelligence agents, but very few have rolled them out broadly across the business. McKinsey’s 2025 global survey found 62% were experimenting with agents, while no more than 10% reported scaling them in any single business function. (forbes.com) The same McKinsey survey found 88% of organizations were using artificial intelligence in at least one business function, up from 78% in 2024. Even with that jump, nearly two-thirds said they had not yet begun scaling artificial intelligence across the enterprise. (welcome.ai) An artificial intelligence agent is software that can plan steps, call tools, and complete tasks with limited human input, not just answer a prompt. McKinsey said agent use was most common in information technology and knowledge management, but most deployments were still confined to one or two functions. (searchyour.ai) The bottleneck is usually not the model. McKinsey’s 2025 findings, as summarized by multiple outlets, pointed to weak data quality, fragmented systems, and limited workflow redesign as the main reasons pilots stall before production. (technologyreview.com) That pattern has shown up in other surveys. Deloitte said in its January 2025 enterprise report that companies were learning they had to bridge the gap between artificial intelligence capability and “operational reality” to get durable results at scale. (deloitte.com) PwC found the same split between enthusiasm and reach. In a May 2025 survey of 300 senior executives in the United States, 79% said artificial intelligence agents were already being adopted in their companies, but two-thirds said fewer than half of employees interacted with agents in daily work. (pwc.com) Boston Consulting Group reported a similar value gap at the budget level. Its AI Radar 2025 survey of 1,803 C-level executives found one in three companies planned to spend more than $25 million on artificial intelligence in 2025, but only 25% said they were seeing significant value. (web-assets.bcg.com) McKinsey’s own benchmark for meaningful payoff was also narrow. Only 39% of respondents said artificial intelligence had any earnings-before-interest-and-taxes impact at the enterprise level, and just 6% qualified as “high performers” with at least 5% earnings impact and significant value from use. (searchyour.ai) The companies getting further were not just adding chatbots to old processes. McKinsey said higher performers were more likely to redesign workflows, set goals tied to growth and innovation rather than only efficiency, and put senior leaders directly behind deployment. (welcome.ai) The result is a familiar enterprise pattern in a new technology cycle: prototypes spread fast, but production systems move only as fast as the data, controls, and operating model underneath them. That is why the headline number is not how many companies tried agents, but how few have made them routine. (forbes.com)
