When PYMNTS Intelligence first began tracking attitudes toward and usage of agentic AI, we found a lot of intrigue from the C-suite, but even more caution. In May 2025, 85% of all CFOs we surveyed said they had no plans to implement agentic AI, leaving just 15% to report actual usage.
Fast forward to August 2025, and it’s a different story and a different C-Suite group checking in on agentic AI. In the time between the two surveys, the industry has seen several key players add the infrastructure needed to make agentic AI ready for the holiday season. Among them: Visa, Mastercard, PayPal, OpenAI and Microsoft. Maybe that promise of stable technology is starting to land, because the C-suite is increasingly looking closer at agentic AI.
This time, we focused on chief product officers. Unlike finance departments, where agentic AI has well-defined use cases, such as automating payables and accounting, teams focused on the product side of enterprises need a degree of imagination to conceive of its potential applications. From a standing start, interest in its potential surged in a mere two months, from June to August 2025. This change reflects a growing recognition of the business case for adopting autonomous AI agents.
“Autonomous” is very much the name of the game, however. Enterprises that already rely on automated systems with varying degrees of autonomy are the most open to adopting agentic AI, which takes generative AI one step further by proactively generating responses and undertaking actions to achieve pre-set goals without constant oversight by humans. In other words, business cultures and processes that value the power of automation and the degrees of autonomy it gives team members are better suited to autonomous technology that takes those cultures and processes to a different level. A company’s current levels of automation and openness to agentic AI correlate highly with its industry segment. Product departments at technology companies report greater interest in agentic AI than those in the service sector or hard goods.
Even within enterprises with a high level of automated systems, however, product leaders remain cautious about giving autonomous agents unfettered access to core systems. This reflects deep concerns over control, security and governance. Trust in the technology is a bigger barrier to adoption than the technology itself.
These are just some of the findings detailed in this edition of the CAIO Report, “From Zero to Beta: How Agentic AI Just Entered the Enterprise Fast Lane,” a PYMNTS Intelligence exclusive report. This edition examines the rise of agentic AI at U.S.-based enterprises. It draws on insights from a survey of 60 CPOs working at U.S. firms that generated at least $1 billion in revenues last year. The survey was conducted from Aug. 7, 2025, to Aug. 22, 2025.
Familiarity With Autonomous Systems Is a Signal for Agentic AI Adoption
The next leg of enterprise automation will not be rules running in the background. It will be software that can independently plan, act and learn inside business systems without human hands. For product teams, the agentic AI promise is tangible: faster design cycles, continuous user testing and sharper competitive positioning. But the payoff is uneven. This latest research shows that agentic AI is starting to take root, but only in organizations already comfortable letting machines make decisions without constant or heavy human supervision. The result is a two-speed digital economy. Enterprises with automated plumbing are moving ahead. At the lower speed, everyone else is watching and worrying about relinquishing human control.
In general, our research maps how agentic AI is entering enterprise product workflows. But adoption is concentrated among companies that already run highly automated systems. Most product leaders buy solutions from vendors rather than building them, and trust is the main brake on progress. Specifically, this highlights a reluctance to grant AI agents action-level access to core systems. In short, readiness and governance, more than the model technology itself, will determine who benefits first.
But we also found that definitions are important in this new and rapidly growing technology. Degrees of prior adoption of automated systems in enterprises can be thought of as degrees of automation in automobiles. The lowest level of automation would be something like cruise control, where one system—acceleration—is automated but everything else, from steering to braking, requires manual control. The highest level of automation would be fully self-driving cars, where the entire operation of the vehicle is autonomous. In enterprises, the lowest level of automation would be like using QuickBooks for accounting. The highest level may be internal use of gen AI to identify new opportunities and reduce fraud.
When it comes to product-related activity in enterprises, automation predicts autonomy. By August 2025, one quarter (25%) of enterprise product departments with the highest adoption of automated systems were also using agentic AI. Another 25% were exploring the adoption of agentic AI within the next 12 months.
Here, half of the surveyed enterprises that were already familiar with and using automated systems had also adopted or were well on their way to adopting agentic AI. These companies are akin to drivers familiar with basic levels of automated systems—say, hands-free assisted parking—who now might be more open to embracing fully self-driving cars.
In contrast, no enterprise product leaders with medium or low prior adoption of automated systems reported adoption of agentic AI, even by August. A handful (3.6%) of those in the medium group were at least piloting autonomous agents.
The findings suggest not just a greater openness to adopting agentic AI among already highly automated enterprises, but that enterprises must reach a certain threshold of adoption of automated systems to be capable of implementing autonomous agents.
Enterprises with the highest current levels of automated systems and the highest degrees of agentic AI adoption tend to be in the tech sector. In August, 20% of tech companies were using or actively exploring adopting agentic AI within the next year. In contrast, only a handful (4.3%) of enterprises in the services sector were even piloting agentic AI endeavors. None had fully deployed these projects. Only 7.4% of companies in the goods sector were exploring adopting agentic AI in the next 12 months. None of these companies had fully deployed or even begun pilot programs.
Notably, interest in agentic AI has surged, even among enterprises with medium or low current levels of automated systems. The share of enterprises with medium levels of automation that report no current plans to adopt agentic AI fell from 90% in June to less than 40% in August. Those with the lowest levels saw a drop from roughly 90% to 50%.
Similarly, in June, all respondents from the services sector reported no current plans to adopt agentic AI. That number fell to just 30% by August. Among goods companies, the ‘no-current-plans’ number fell from 96% in June to 56% in August.
Product Leaders Favor Buy Over Build
More than 90% of product leaders lean on outside vendors or consultants to help chart their path into agentic AI. However, there is some variation among industry segments. All service-sector enterprises rely on third parties. Most tech companies, at 80%, rely on third parties, perhaps reflecting a difference in levels of in-house expertise.
There are also notable differences in how enterprises in different segments anticipate implementing agentic AI systems. Only 8.3% of product leaders at goods companies are training staff to work alongside agentic AI systems, suggesting that they plan to continue relying on vendors and consultants for agentic AI-related work. In contrast, 25% of product leaders in service companies and 20% in tech companies are training staff to work with agentic AI. This shows their greater interest in bringing agentic AI operations in-house.
Training staff takes time and can be expensive, but it can lead to greater control and greater in-house expertise, which can drive efficiency in the long run. On the other hand, working with outside vendors can enable enterprises to move quickly with a lower up-front investment. The differences across industry sectors may reflect differences in how much value enterprises in different sectors envision agentic AI adding to their business.
A similar pattern was found with embedding agentic AI into existing enterprise software. Just 8.3% of goods company product leaders anticipate embedding agentic AI into existing systems. At the same time, 20% of tech company product leaders anticipate doing so.
Enterprises Are Not Adopting Agentic AI Uniformly
Product teams across different industry segments have different preferred use cases for agentic AI. Those reflect the different pain points and value drivers for each particular industry. The most common preferred use case for goods and tech companies is product ideation and design. One-third of respondents in each sector cited this.
Beyond product innovation, different industry segments have different priorities for agentic. Tech-sector product leaders split their responses evenly into three use cases: product ideation and design, user and accessibility testing, and project lifecycle management. In contrast, service enterprises spread their responses across all six use cases tested.
For tech companies, user experience is a key competitive differentiator. Continuously testing that experience, therefore, is a high-value add use case for agentic AI. User experience is also a key differentiator for many service-sector businesses, which is likely why it ranked higher than lifecycle management or competitive analysis.
Goods sector product leaders were more focused on using agentic AI for competitive analysis. That reflects the need to stay ahead in product development cycles and market positioning with respect to rivals. Using autonomous agents to continuously monitor competitors’ pricing, merchandising strategies and new product additions is therefore a high-value-add use case in that industry.
Trust is the Bottleneck
Despite the surge in interest in adopting agentic AI, product leaders across all industry segments and all levels of prior adoption of automated systems remain cautious about giving AI agents access to core internal systems. Overall, 98% say they are not at all ready to grant core-system access to fully autonomous agents.
Trust, then, even more than the technology itself, is a major constraining factor on adoption. Even among enterprises with high prior levels of automation, three out of four (75%) have trust issues with agentic AI. Issues of security and compliance are still regarded as too risky to be left to fully autonomous systems without human oversight.
The implications for vendors are clear. To accelerate adoption and expand the range of applications for agentic AI within enterprises, providers must focus on demonstrating concrete measures of transparency, auditability and security. Without those, agentic AI will remain a hard sell into enterprise product departments.
Read More
PYMNTS Intelligence is the leading provider of information on the consumer trends driving innovation in consumer finance, digital payments and financial inclusion. To stay up to date, subscribe to our newsletters and read our in-depth reports.
Methodology
“From Zero to Beta: How Agentic AI Just Entered the Enterprise Fast Lane” is based on a survey of 60 CPOs working at U.S. firms that generated at least $1 billion in revenues last year. The survey was conducted from Aug. 7, 2025, to Aug. 22, 2025, and we compared the results with those from a similar survey in June 2025. The report examines the growing openness to agentic AI among enterprise CPOs. The sample includes executives with varying levels of strategic authority, enabling segmented analysis by sector, confidence level and risk posture. PYMNTS Intelligence designed the survey to assess expectations for generative and agentic AI performance across use cases, and how those expectations have evolved year over year.
