New Delhi, July 02, 2025: Nasscom today released its latest research report, “Enterprise Experiments with AI Agents – 2025 Global Trends,” offering a first-of-its-kind comprehensive lens into how global enterprises are experimenting with and investing in AI agents. At a time when AI is shifting from passive analytics to active agents of execution, the report captures the foundational shifts underway within the enterprise ecosystem. The study draws on responses from over 100 global enterprises across 8 -9 major regions and over 10 industries, offering a landscape view into how businesses are transitioning from early stage Generative AI applications toward more goal-oriented, human-plus-AI agentic systems.
Enterprises are strengthening their tech core from AI spend and talent, GenAI capabilities, data foundations, and process flexibility to build and deploy AI agents. The study underscores that 88% of enterprises now have dedicated AI budgets, with two-thirds of them allocating over 15% of their tech budgets specifically toward AI initiatives. This shift is reflected in the emergence of specialized AI teams, greater focus on GenAI platforms and tooling, and infrastructure readiness. However, while awareness of GenAI is high, actual use of advanced models remains limited. Only about half of surveyed enterprises are fine-tuning large language models (LLMs) or foundation models for their own applications.
Crucially, the move toward Agentic AI is gaining traction. Nearly 62% of global enterprises are currently experimenting with such AI agents, ranging from proof-of-concepts to scaled pilots. However, the nature of these experiments is still largely internal, focused on task-level automation with human oversight, with 76% of enterprises positioning their own IT operations as “client zero.” External-facing use cases, such as customer service, are still limited, with only 31% of companies indicating active usage in those areas. Yet, a significant 88% of enterprises indicate intent to dedicate specific AI budgets toward agentic systems in 2025, suggesting strong optimism but cautious execution.
Sangeeta Gupta, Senior Vice President and Chief Strategy Officer at Nasscom, said, “We are at the tipping point of the AI maturity curve where enterprises are no longer just experimenting with AI, but actively reimagining their architecture, workflows, and teams to build agentic systems. AI agents represent the next evolution of enterprise AI one that requires philosophical shifts in how we view work, intelligence, and autonomy. But to scale responsibly, trust, data readiness, and human oversight will be non-negotiable.”
Despite growing confidence, the study reveals that deployment remains largely incremental. A significant 77% of enterprises are adopting agentic AI systems with a “human-in-the-loop” design, reflecting an awareness of the need for constant oversight, adaptability, and contextual judgment. While 46% report experimenting with autonomous agents. IT operations, customer service, and internal HR and finance functions are leading experimentation grounds. Manufacturing enterprises are moving faster than services in adoption, with AI-powered robotics, quality control, and process agents showing strong traction.
The business case for Agentic AI appears strongest in real-time decision-making and operational agility. More than half the enterprises see such systems as critical enablers for translating information into intelligence and rapidly responding to shifting market dynamics. Only 39% believe that agentic systems will meaningfully free up human bandwidth for higher-order work, suggesting that, at present, these systems may augment rather than replace existing workflows.
Data remains the cornerstone of AI efficacy. With 68% of companies focusing on strengthening data governance and management, and 62% working on integrating structured and unstructured data flows, enterprises are laying the groundwork for scalable, reliable agentic solutions.
However, the path forward is marked by both technical and structural headwinds. Data privacy, risks of self-learning systems, and the absence of cohesive regulatory frameworks continue to be cited as top adoption barriers. While most enterprises still rely on adapted legacy risk frameworks, only 43% enterprises have initiated focused AI risk protocols, including observability tools and hardware-level audits. Interestingly, only 44% of companies expressed concern about the cultural and mindset shifts
required to build effective human + AI systems or the perceived limitations in ROI from such deployments. While just 27% of global enterprises identified the lack of AI talent as a major constraint.
However, success will not be defined by autonomy alone. Enterprises need to prioritize human-AI collaboration, ensure process adaptability, and embed trust at the core of their systems to lead in the age of intelligent agents.
