Enterprises wary of agentic AI due to data, talent and RoI concerns

While many enterprises are exploring AI agents for active decision making and complex problem-solving purposes, these efforts often fail to deliver their full potential, according to a research report published by the IT industry’s apex body, Nasscom, on Monday.

Agentic AI are intelligent systems, capable of independent decision-making and learning, with CXOs often acknowledging how these can autonomously make decisions. In reality, however, adoption faces hurdles due to rising costs, unclear business value, data privacy, insufficient risk controls and a lack of cohesive regulations.

The report titled, “Enterprise Experiments with AI Agents – 2025 Global Trends,” noted that most enterprises still rely on adapted legacy risk frameworks, only very few focused on AI risk protocols, including observability tools and hardware-level audits. Furthermore, companies expressed concern about the cultural and mindset shifts required to build effective human plus AI systems or the perceived limitations in ROI from such deployments. Enterprises also identified the lack of AI talent as a major constraint.

To be sure, a study released by analyst firm Gartner noted similar concerns, predicting that more than 40% of agentic AI projects will be canceled by the end of 2027. Many of the current agentic AI projects, according to Anushree Verma, Senior Director Analyst at Gartner, are hype-driven early experiments that are often misused. Verma states, “This can blind organisations to the real cost and complexity of deploying AI agents at scale, stalling projects from moving into production,” she said.

Likewise, Arun Hiremath, co-founder and chief business officer at AI-powered automation platform EvoluteIQ, said, “Transparency and explainability are also crucial—AI decisions shouldn’t be a black box. Organisations need clear mechanisms for understanding and tracking AI reasoning to ensure alignment with business goals.”

Additionally, human oversight remains important. While Agentic AI excels at autonomy, high-stakes or complex decisions still require human judgment and escalation paths. Bias and fairness are also critical concerns; enterprises must conduct continuous audits to prevent discrimination and ensure ethical outcomes, noted reseatchers.

The other caveat is that the lack of quality data often acts as a barrier to effective AI implementation. “If the data is insufficient, incomplete, or inaccurate, the resulting AI will produce subpar or incorrect results. Hence, the effectiveness of AI models in your business is fundamentally dependent on having suitable data that is well-organised, clean, and thorough,” Debo Dutta, Chief AI Officer of Nutanix, said.

Therefore, he believes that, before hastily integrating AI into business processes, it is crucial to ensure the correctness of the data. 

Despite these challenges, Agentic AI is gaining traction, with nearly 62% of global enterprises experimenting with it, primarily for internal task automation with human oversight. External-facing applications, such as customer service, remain limited.

Sangeeta Gupta, Senior Vice President and Chief Strategy Officer at Nasscom, observed that enterprises are actively reimagining their architecture and teams to build agentic systems, marking a key shift in how we view work and autonomy. She believes that trust, data readiness, and human oversight are essential for responsible scaling of the technology.

Ultimately, experts believe that prioritising human-AI collaboration, ensuring process adaptability, and embedding trust are critical for success in the age of intelligent agents.