84% See AI Productivity Gains, but Only 10% Allow Agent Autonomy: TCS Study

Tata Consultancy Services (TCS) found that 84% of executives across enterprises identified productivity improvements as the leading benefit of AI investments. However, only 10% of enterprises allow AI agents to execute decisions without human approval across multiple business functions. 

Only 9% of executives reported that their organisations had AI agents that could securely orchestrate workflows across enterprise applications.

Security, risk and governance emerged as the leading considerations when organisations decide where and how to deploy AI systems. These factors ranked ahead of cost, speed and ease of deployment.

Amit Kapur, Chief AI and Services Transformation Officer at TCS, said, “What seems to be holding organisations back – and rightly so – are concerns around security, governance, and resilience.”

He added, “Deploying AI responsibly, with the right guardrails and enterprise controls, will be the big focus in the next stage of this journey.”

As agents gain access to enterprise systems, observability becomes important for monitoring their actions and decisions.

The study also found that 90% of respondents considered model independence an important factor in AI investment decisions. However, only 7% reported having organisation-wide continuity plans to switch AI providers.

About 34% of respondents identified AI sovereignty requirements, including control over data location, model choice, and regulatory compliance, as a high or very high organisational risk. In Europe, the figure rose to 43%.

Separately, legacy systems, fragmented data and integration gaps are emerging as obstacles to scaling AI deployments. A recent assessment of Indian enterprises highlighted infrastructure challenges in moving AI workloads into production, as organisations assess whether their existing infrastructure can support AI at scale.

AI economics remain unsettled. Thirty-nine percent of enterprises said they were still assessing whether their AI initiatives generated sufficient value relative to token and compute costs.

Fidelity International has introduced use-case-level token budgets and an internal AI gateway to monitor expenditure and restrict access. Its approach to managing AI costs and model access offers an example of how organisations are addressing these concerns.

Meanwhile, 44% of respondents said important business knowledge remained embedded in employees' expertise or fragmented documentation, potentially limiting AI systems' access to enterprise context. Another 46% reported deploying open-source AI for selected use cases.

The proportion of leaders who said it was too early to determine AI's effect on jobs rose from 4% in 2024 to 19% in 2026.

North America showed particular strength across several enterprise AI foundations, with 15% of respondents treating AI as a strategic enterprise capability. The UK and Ireland reported comparatively higher levels of continuity planning and governance integration.

The study, titled Intelligence that is Real, surveyed 1,333 senior business and technology leaders across 25 countries and 12 industries. Their organisations had annual revenues ranging from $500 million to more than $20 billion.