論説
概要
- AI is becoming more important to companies: 74% say it’s a top-three strategic priority (vs. 60% a year earlier), and 21% call it their top priority.
- The perceived risk of disruption from AI is increasing, especially in tech, where 44% of companies see high or very high disruption risk.
- Top use cases are moving to scale: 40% in software development, 32% in customer service, and more than 20% in many other domains.
- Results are real, though not a given: Executives say that 80% of generative AI use cases met or exceeded expectations, but only 23% can tie initiatives to new revenue or lower costs.
AI remains a top priority for most companies, and many are beginning to see business results from their scaled efforts. Between the third quarter of 2024 and the third quarter of 2025, the percentage of companies ranking AI as a top-three strategic priority rose from 60% to 74%. The share who ranked it as their No. 1 priority more than doubled, with 21% of respondents ranking it on top.
Disruption risk. At the same time, concerns are growing about the risks that AI brings to some industries. The percentage of companies saying that AI posed a very high risk of disruption in their industry more than doubled between the fourth quarter of 2024 and the third quarter of 2025—the same period when agentic AI became the focus of discussion and experimentation.
This tension is more evident in the technology sector, which is likely to see greater disruption sooner. About 17% of tech companies now see AI posing a very high risk of disruption to their industries, and 44% see a high or very high risk, compared with 5% and 36%, respectively, for companies in other sectors. Across sectors, a majority see at least a moderate risk of disruption.
Across domains. Companies are also extending their use of AI into more functions and domains—far more rapidly in just three years than anything seen in previous technology waves. For example, 73% said they use AI in software development, up from 66% a year earlier. Similar percentage increases were reported in customer service, knowledge worker efficiency, marketing, IT, and other domains. And even where adoption rates are lower, growth is rapid.
Scale to production. These adoption levels produce a counternarrative to the popular misconception that AI deployment doesn’t reach beyond the pilot stage. Most use case categories are seeing an increase in the percentage of pilots moving to production at scale. Software development is a clear leader, with 40% of pilots moving to scale—a good indicator that it fits well with AI capabilities. There’s also a solid second tier of domains where between one-fifth and one-third of use cases are scaling, including customer service, sales, marketing, and knowledge worker efficiency.
Roadblocks. Many of the concerns that have held back adoption of generative AI have begun to see a very gradual reduction, including those related to in-house expertise, quality and accuracy, return on investment, and data readiness. The notable exception remains concerns around data security and privacy, which have risen over the past year, especially among companies that have moved from pilots to production.
Business results. As companies move AI out of pilots and into production, most report greater satisfaction with the results. Our survey found that of the 59% of companies that are meaningfully adopting generative AI, the technology met or exceeded expectations in about 80% of cases across domains. About 62% of those who said generative AI met or exceeded expectations also cited improved business results or successful transformation due to its deployment, and 78% of those respondents said these results resulted in measurable revenue increases or cost decreases. In all, about 23% of all respondents said the use of generative AI had delivered more revenue or lower costs.
Satisfaction with automation. Interestingly, satisfaction appears to increase as companies progress from using AI as an assistant to assigning it task or agentic workflow automation. Respondents using AI for agentic workflow automation were twice as likely to say it exceeded goals as those who use it as an assistant, and only half as likely to report disappointment.
Disappointments. Where AI failed to meet expectations, companies told us the technology could address some work tasks but not others. About 33% of these unsatisfied respondents said the technology worked at the pilot level but didn’t scale. About the same percentage said it was more expensive to develop than anticipated.
Three years after generative AI began its ascent as an essential business technology, executives across industries show few signs of losing interest. More companies are rating AI as a top priority, and they are getting more serious about developing strategies and managing effective deployment. The public dialogue periodically raises doubts about the success of AI deployment or the likelihood of significant return on investment. But our recurring surveys find that companies are deploying AI beyond pilots at scale, achieving high levels of satisfaction, and delivering real business results—though they are not a given. Indeed, the rapid embrace of AI exceeds the pace of any technology we’ve seen yet and is likely to persist as long as businesses continue to find innovative and productive applications for AI.
Survey: Generative AI’s Uptake Is Unprecedented Despite Roadblocks
Bain’s periodic assessment on generative AI readiness finds nearly ubiquitous adoption, tempered by security and quality concerns.
First published in 11月 2025