Only 10% of Enterprises Let AI Agents Decide Without Human Approval
A TCS survey of 1,333 leaders across 25 countries finds 84% cite AI productivity gains, yet only 10% let AI agents decide without human approval across functions.

Why it matters
- 84% of executives rank productivity as the main benefit of AI investment, per a TCS study of 1,333 leaders across 25 countries published November 17, 2026.
- Only 10% of organisations allow AI agents to make decisions without human approval across multiple functions; only 9% have agents that securely coordinate workflows across applications.
- 90% of respondents call model independence important in AI investment decisions, but only 7% have provider-switching continuity plans in place.
- 43% of European respondents rate AI sovereignty-related requirements as a high or very high organisational risk, versus 34% globally.
- Leaders saying it is too early to determine AI's workforce impact rose to 19% in 2026 from 4% in 2024.
Eighty-four percent of executives rank productivity improvement as the main benefit of their AI investments, yet only 10% of organisations let AI agents make decisions without human approval across multiple business functions. That is the central tension in a global study on enterprise AI adoption published by Tata Consultancy Services on November 17, 2026. Companies are extracting real value from AI-assisted work — but they are keeping their hands firmly on the wheel.
The gap matters because the AI industry's next pitch to enterprises is autonomy. Vendors want AI agents that don't just draft text or summarise documents, but that coordinate workflows across internal applications and act on their own. The TCS data suggests most buyers are not ready for that step, and the reasons are structural rather than sentimental.
What did TCS actually measure?
The study surveyed 1,333 senior business and technology leaders across 25 countries and 12 industries. Respondents included C-suite executives, divisional leaders, senior vice presidents and vice presidents at organisations with annual revenue from USD $500 million to more than USD $20 billion — in other words, the large enterprises vendors court hardest.
Alongside the headline numbers on autonomy, the survey found that only 9% of organisations have AI agents that can securely coordinate workflows across enterprise applications. Security, risk and governance emerged as the dominant factors shaping how businesses deploy AI systems, ranking ahead of cost, speed and ease of deployment. The findings suggest many companies are moving deliberately as AI tools gain access to internal applications, workflows and decision-making processes.
Amit Kapur, Chief AI and Services Transformation Officer at TCS, framed the results as a story of progress with a hard brake. "We are seeing significant progress in scaling AI for enterprises," Kapur said. "AI is clearly moving from assisting people to taking more autonomous action with deep access into an enterprise."
Kapur noted that productivity is where the value currently concentrates. "Productivity is where enterprises are seeing the greatest value from AI today, with 84% of executives ranking it as the leading source of value from their AI investments, though 10% also said they have seen a revenue uplift," he said. On the restraint, he was blunt: "What seems to be holding organisations back — and rightly so — are concerns around security, governance and resilience. Deploying AI responsibly, with the right guardrails and enterprise controls, will be the big focus in the next stage of this journey."
Why is the autonomy number so low?
The study points to a readiness gap rather than a demand gap. Businesses are finding immediate value through productivity gains, but relatively few have the controls, technical integration and risk frameworks needed to support more autonomous use. An agent that executes decisions across functions requires governance that most enterprises have not yet built.
That caution also shows up in how companies plan around model dependence. Ninety percent of respondents said model independence is an important factor in AI investment decisions — a signal of widespread concern about becoming tied to a single provider or technical approach. But recognising the risk and preparing for it are different things:
- Only 7% of organisations have provider-switching continuity plans in place across the business.
- 34% of all respondents identified AI sovereignty-related requirements — control over data location, model choice and compliance — as a high or very high organisational risk.
- In Europe, that sovereignty concern runs stronger, at 43%.
The regional detail cuts both ways. The study found the UK and Ireland stood out for relatively strong continuity planning and governance integration, while North America showed strength across several enterprise AI foundations.
Does AI pay for itself?
For many enterprises, that question is still open. The survey found 39% of enterprises are still assessing whether AI initiatives generate enough value relative to token and compute costs. That is a striking figure for technology that has been deployed at scale for two-plus years at many large firms, and it suggests the economics of inference remain unresolved for a substantial share of the market.
Access to usable knowledge is a second bottleneck. Forty-four percent of respondents said important business knowledge remains embedded in employees or scattered across fragmented documentation. That fragmentation limits AI systems' ability to draw on consistent enterprise context — an agent can only act on what it can reliably read.
On the technology side, open-source AI is becoming more common in selective deployments. Forty-six percent of respondents reported targeted use of open-source models for specific applications, a sign that the proprietary-model default is eroding at the edges, in line with the model-independence findings.
What about jobs?
Uncertainty is growing, not shrinking. The proportion of leaders who said it is too early to determine AI's impact on the workforce rose to 19% in 2026, up from 4% in 2024. That five-fold jump suggests business leaders remain unsure not only about the scale of workplace change, but also how quickly it will happen and which roles will be most affected.
Productivity remains the leading source of value, the study found, followed by improvements in customer experience and the development of new products or services.
The research frames the broader shift as a move toward what TCS calls a Human+AI operating model — one in which AI systems progress from supporting employees to taking on more responsibility in selected settings. The framing is deliberate: it positions autonomy as a staged rollout, not a switch that flips.
What happens next?
For many organisations, the study concludes, the next step is less about proving AI can deliver value and more about deciding where oversight must remain with people and where systems can be trusted to act. That is a governance question, not a capability question — and with only 10% of enterprises granting cross-functional autonomy today, it is the question that will define enterprise AI adoption through 2027.
Source: itbrief.com.au
More from Elena Vasquez
Show full bio
Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.
708 articles
Related articles
- Productivity leads enterprise AI, but autonomy remains scarce
- 51% of AI projects miss targets, yet 85% of firms raise budgets
- ANZ AI investment runs ahead of data readiness and governance
- AI Agents Become Coworkers—and Managers of the Bots
- SAS: Corporate AI's 2027 Test Will Be Trust, Not Deployment Speed