In the meantime, the survey of 214 tech leaders across companies reveals that 38% are still experimenting through isolated pilots and another 32% have deployed AI in select use cases.
“While cost is a significant factor, the slow transition from pilot to scale is primarily driven by broader operational and strategic hurdles,” said Arun Shetty, chief technology officer, and senior director, solutions engineering, Cisco India and South Asia.
Organisations face challenges in centralising data, securing sufficient GPU capacity, and overcoming data bottlenecks. Many lack repeatable frameworks to move from pilots to production, he said.
Infrastructure modernisation
The survey findings highlight a widening gap between enterprise AI ambitions and operational readiness, with respondents identifying data management, infrastructure modernisation and governance as critical bottlenecks. “Infrastructure constraints, trust, and data readiness remain the biggest barriers to AI adoption,” Shetty said.
As organisations move from experimentation to deployment, rising demand for power, compute, storage, and networking resources is straining existing infrastructure, he added.The survey also pointed to the importance of infrastructure capabilities over the next three to five years. Nearly nine in ten respondents described real-time data processing and hybrid cloud connectivity as either “very important” or “extremely important” for their AI strategies.
Serious intent
“Indian enterprises are taking AI seriously, and the survey captures that intent with a fair amount of clarity,” said Saurabh Saxena, vice president, India, OpenText, a company that specialises in information management. “What stands out is the gap between ambition and actual readiness. Only 5% of organisations say AI is already part of their core operations, while most remain in pilots or limited deployments.”
Ankush Tiwari, chief executive of pi-labs, told ET, “Nearly 80% of AI adoption today is relatively frictionless. It is the remaining 20%, making AI deterministic, reliable, auditable and enterprise-grade at scale, which is consuming disproportionate time, investment and operational patience.”
Data quality
Vikram Raichura, chief executive of Helo.ai, said organisations that have successfully deployed AI tend to focus first on governance, compliance and data quality before investing in advanced models.
Industry executives say companies face difficulty in securing AI applications and agents against evolving AI-specific threats. Moving beyond tactical use cases requires a fundamental shift in business models, which many organisations are still navigating. The primary barrier remains the complexity of operationalising AI at scale. As AI becomes more pervasive, “you can’t trust what you can’t see” becomes increasingly important, making end-to-end visibility across the AI stack essential, according to Cisco’s Shetty. This trust deficit mandates protecting AI itself, making safety and security prerequisites for deployment and operation. Without well-managed, accessible, high-quality data, AI’s potential remains untapped, he said.
(This article is part of the AI Vantage series, developed in partnership with Cisco)