In an intriguing revelation from a recent Harvard Business Review Analytic Services research report, it has been found that a mere 6% of companies are comfortable with AI agents autonomously managing their core business processes. This statistic reflects a significant chasm between the enthusiastic embrace of artificial intelligence and the cautious approach many organizations are taking regarding their most critical workflows.
The report, sponsored by automation platform provider Workato and Amazon Web Services, surveyed 603 business and technology leaders worldwide in July 2025. The findings indicate that although agentic AI—intelligent systems capable of decision-making and action with minimal human oversight—is proliferating at remarkable speed, trust has not kept pace, especially in high-stakes environments.
Interestingly, while 43% of respondents express trust in AI agents for limited or routine operational tasks, 39% admit they restrict these agents to supervised scenarios or non-core processes. This pattern illustrates a predominant cautious stance: businesses exhibit a willingness to experiment with AI, yet remain hesitant to unleash it fully on decisions that could affect finances, customers, or the workforce.
Adoption Outpaces Readiness
Despite the reluctance to fully trust AI, organizational deployment is burgeoning. Nine percent of organizations report they have completely deployed agentic AI, while half are actively piloting or exploring various use cases. Strikingly, only 10% of these organizations have opted out after initial consideration. Notably, 86% of respondents expect investments in agentic AI to rise over the next two years, indicating a commitment to deepening experimentation even amid prevailing skepticism.
However, readiness to support such adoption lags alarmingly behind. Just 20% of participants believe their technological infrastructure is adequately prepared for agentic AI in core operations, while only 15% rated their data and systems as fully ready. A concerning 12% feel that risk and governance controls are effectively in place. The report defines organizational readiness using a composite index across various areas including infrastructure, data, cybersecurity, and governance, categorizing organizations into “leaders” (27%), “followers” (50%), and “laggards” (24%).
Benefits Real, but Below Expectations
Early adopters of agentic AI are experiencing actual gains, yet many have not realized the full benefits they anticipated. Most organizations deploying or piloting this technology report improved productivity, reduced costs, and enhanced customer experience as primary gains. However, the outcomes often fall short of expectations. This underscores the necessity for robust foundations in data quality, governance, and technological architecture, as a lack of these can lead to “garbage in, garbage out” scenarios that could ultimately dismantle trust in AI.
Concerns over security and privacy pose significant roadblocks to broader AI adoption. About 31% of respondents identify cybersecurity and privacy anxieties as a major challenge, followed closely by worries over the quality of data outputs (23%), unprepared business processes (22%), and technology infrastructure deficits (22%). Organizations are also grappling with governance and regulatory uncertainties, further reinforcing the tendency to limit AI agents to less critical tasks.
In light of these issues, many companies are turning to an approach termed “enterprise orchestration.” This strategy promotes the integration of systems, data, and applications into a governed layer, offering a secure foundation for deploying AI agents at scale. Currently, 8% of organizations report having implemented enterprise orchestration in anticipation of agentic AI, while 74% are either actively working on it or planning its implementation. More than 80% consider connecting applications and accessing contextual data critical to the success of these orchestration efforts.
The Human Factor and the Road Ahead
The report accentuates that technology alone isn’t sufficient; the human factor is equally, if not more, significant in determining whether AI agents can transition from peripheral roles to core enterprise functions. Kim Huffman, chief information officer at Workiva, emphasizes the need for effective change management and reskilling, indicating that this aspect has frequently been undervalued in AI initiatives.
A substantial 44% of organizations prioritize training or upskilling employees to oversee agentic AI, while 39% are shifting focus toward developing responsible AI governance frameworks. Some companies are even appointing AI ambassadors and champions within various functions to identify potential use cases and guide teams through preliminary piloting stages.
Though only 6% currently fully trust AI agents with core processes, 72% of respondents assert that the overall benefits of agentic AI outweigh associated risks. Leaders in this field express the most confidence in this trade-off. The study suggests that as organizations enhance investments in orchestration, governance, and workforce training, the troubling trust gap may diminish, potentially allowing AI agents to evolve from experimental aides to trusted overseers of workflows integral to corporate success. This journey will be pivotal as enterprises navigate the complexities of AI integration into essential business practices.
For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.
