93% of IT execs are eager to implement agentic AI – but have they considered governance?

In the last two years, we’ve seen a massive demand for AI technologies, particularly with the rise of GenAI, but their value for enterprises as standalone capabilities has plateaued. Companies need new, more robust ways to leverage AI that can boost their efficiency and productivity so they achieve a more significant return on their AI investments.

Recent data indicates that businesses across all industries are excited about the potential of AI agents, software robots that can use AI skills to accomplish more complex tasks, to do just that. But what happens when AI agents proliferate? Will companies design and govern them properly?

According to a recent survey from independent authors and UiPath of over 250 U.S. IT executives, 93% of respondents are extremely or very interested in agentic AI, with 37% noting that they already use the technology. Additionally, 90% of IT executives say they have business processes that agentic AI would improve.

Because AI agents can learn and improve over time, use cases are nearly limitless, but the report found current interest fits into a few distinct buckets. Most of the executives surveyed view AI agents as a path to improving oversight and business workflows (58%), increasing application integration (53%), and automating complex business workflows (52%).

The potential benefit of AI agents to augment and automate everyday end-to-end workflows is clear. Yet, despite the excitement and the success of early adopters, concerns over security and governance could slow near-term implementations of AI agents.

Controlled agency to deal with security concerns of AI agents 

The most common concern by US IT executives in the report is IT security issues, cited by 56% of respondents. AI agents derive much of their value from the ability to gain context from company data and act on that information. That important capability also creates risk, especially as AI agents become increasingly interconnected with enterprise applications.

AI agents will not replace everything but, instead, they will be an essential part of a broader orchestration of workflows and processes. Thereby, controlled agency will ensure that AI works in harmony with human decision-making and existing automation workflows.

Controlled agency is a methodology for improving the accuracy, resiliency, and autonomy levels of AI agents. It’s essential for ensuring the trustworthiness of actions and decisions within the agentic orchestration of AI agents, robots, and people. As enterprises scope their agentic AI deployments, controlled agency will prove to be a crucial factor in their success.

So, while security concerns surrounding agentic AI are legitimate, by using agentic automation with controlled agency, organizations can trust how they operate with agents.

With agentic automation, AI agents leverage robots to execute specific actions. By limiting the types of robots that agents have access to, organizations can prevent agents from going rogue and accessing data they should be avoiding. Agents also have an additional layer of oversight from humans, who oversee multiple agents to ensure all of them are working effectively toward organizational goals. There are also additional security measures, like encryption, access controls, and regular audits, to further safeguard data and maintain regulatory compliance.

Under this model, employees control the entire framework, while the agents are still able to work autonomously to make decisions and learn over time to improve their outputs. The employees overseeing these agents and robots benefit from being able to offload more repetitive work onto agents so they can work on more high-level tasks.

Keys to agentic implementation 

In addition to security and governance concerns, IT executives say the top limitation of existing AI tools is a lack of integration with other business applications. Of the executives surveyed, 87% say interoperability between different AI technologies is essential or significant to their organizations.

To overcome integration issues and have different technologies work together seamlessly, enterprises should work with trusted partners and experts who understand these complexities and can streamline the integration process. This will help enterprises navigate the complexities of agentic AI and the nuances of their specific organization to achieve seamless implementation.

It’s important to remember that agentic implementation is not a patch-up project but a long-term journey. Enterprises should take a phased approach to adoption, starting implementation with an internal, smaller-scale process with little impact on finance, cybersecurity, and data privacy. This approach allows organizations to become familiar with different challenges and solutions by integrating AI agents with existing systems, other agents, and digital robots. This way, they can have an area within the organization to champion this framework before building it out to other company areas.

Implementing agentic AI becomes possible with an agentic automation platform that expands an organization’s potential by automating end-to-end workflows and processes. Agentic automation represents a new era in which agents imbued with AI form the core of automated operations. These AI agents exceed mere tools; they are collaborators involved in making decisions, actively learning, and working alongside people and robots to achieve enterprise objectives.

Read the UiPath 2025 Agentic AI Research Report to learn more about how agentic automation can help your organization.