Artificial intelligence is delivering measurable productivity benefits across businesses, but most enterprises remain reluctant to give AI systems the authority to make decisions independently. A new global study by Tata Consultancy Services (TCS) found that 84% of surveyed executives identify productivity improvement as the leading source of value from AI investments. However, only 10% of global enterprises allow AI agents to execute decisions without human approval across multiple business functions.
The findings highlight a growing gap between AI adoption and autonomous execution. While companies are increasingly using AI to support employees, analyse information and automate routine activities, concerns about cybersecurity, governance, operational resilience and accountability are slowing the transition towards independent AI agents. The TCS Enterprise AI Study, titled Intelligence that is Real, suggests that the next phase of enterprise AI will depend less on simply adopting models and more on building the infrastructure and controls needed to trust them.
TCS Study Reveals a Gap Between AI Productivity and Autonomy
The TCS study found that productivity improvement remains the strongest source of value from enterprise AI investments. Companies are increasingly using AI to support knowledge workers, accelerate processes and improve operational efficiency.
However, allowing an AI agent to perform an entire business workflow without human intervention is a different challenge.
Only 10% of surveyed enterprises permit AI agents to execute decisions without human approval across multiple functions or business areas. Even more notably, just 9% of executives said their organisations had AI agents capable of securely orchestrating workflows across enterprise applications.
This distinction is important. An AI assistant can recommend an action or prepare a document, but an autonomous agent may need permission to access company databases, communicate with other applications and execute decisions.
For example, an AI agent working in a financial department could analyse invoices, identify discrepancies and recommend payments. Allowing it to actually approve and execute payments without human review requires a much higher level of security, accountability and operational control.
The study indicates that most organisations are still developing the foundations required for this transition.
Security and Governance Are the Biggest Barriers
Security, risk management and governance emerged as the leading considerations influencing enterprise AI deployment decisions.
These concerns are becoming more important as AI agents receive access to sensitive company information and business applications.
An autonomous agent could potentially access customer records, financial information, internal communications and operational systems. Without appropriate controls, an incorrect decision or compromised agent could create financial, legal or reputational consequences.
TCS Chief AI and Services Transformation Officer Amit Kapur said organisations are making progress in scaling AI, but security, governance and resilience remain important considerations before granting AI systems greater autonomy.
The company believes that responsible deployment will require stronger enterprise controls and safeguards.
These include:
- Access permissions that restrict what AI agents can view or modify.
- Monitoring systems that record agent activities and decisions.
- Human approval for sensitive or high-impact actions.
- Security testing to identify vulnerabilities.
- Governance frameworks that establish accountability for AI-generated decisions.
For businesses, the challenge is finding a balance between automation and human oversight.
90% of Executives Want Model Independence
The study also identified growing concerns about dependence on individual AI providers.
Around 90% of respondents said model independence was an important consideration in their AI investment decisions. This means businesses increasingly want the flexibility to switch between different AI models and providers rather than rely entirely on one supplier.
However, only 7% reported having organisation-wide continuity plans for switching AI providers.
This gap creates a potential business risk. If an AI provider changes its pricing, restricts access, experiences an outage or modifies its model capabilities, organisations without alternative arrangements could face disruption.
Model independence is also connected to the growing importance of AI sovereignty.
About 34% of respondents identified sovereignty-related requirements as a high or very high organisational risk. These requirements include control over data location, model selection and regulatory compliance.
The concern was particularly significant in Europe, where 43% of respondents identified AI sovereignty as a high or very high risk.
For multinational businesses, the issue is becoming increasingly complex because AI systems may process information across different countries and regulatory jurisdictions.
AI Economics Remain Uncertain for Many Companies
Although enterprises are investing heavily in AI, many are still evaluating whether the technology generates sufficient financial returns.
The TCS study found that 39% of enterprises continue to assess whether their AI initiatives deliver enough value relative to token and computing costs.
This is an important challenge for businesses deploying large language models and autonomous agents.
AI systems generate costs through model inference, data processing, cloud infrastructure, application integrations and ongoing monitoring. Autonomous agents can also make repeated model calls while completing multi-step tasks, potentially increasing computing expenditure.
Companies therefore need to measure whether automation produces enough savings or additional business value to justify these expenses.
The study also highlights a gap between productivity improvements and broader financial outcomes. While 84% identify productivity as the leading source of AI value, only 10% of executives reported an uplift in revenue.
This suggests that improving employee efficiency does not automatically translate into higher sales or stronger financial performance.
Businesses may need to redesign workflows, develop new products and services, and connect AI initiatives directly to commercial objectives to realise wider financial benefits.
Fragmented Business Knowledge Limits AI Adoption
Another significant challenge identified by TCS is the availability of reliable enterprise information.
Around 44% of respondents said important business knowledge remains embedded in employees’ expertise or scattered across fragmented documentation.
This creates problems for AI systems that depend on accurate and accessible information.
For example, an AI agent handling customer service may need access to product policies, customer histories, pricing information and internal procedures. If these records are incomplete or stored across disconnected systems, the agent may produce incorrect answers or fail to complete its task.
Businesses therefore need to improve data quality, integration and knowledge management before deploying AI agents across complex operations.
The study also found that 46% of respondents reported targeted deployment of open-source AI for selected use cases, showing that companies are exploring alternatives to relying entirely on proprietary models.
AI’s Impact on Jobs Is Becoming Less Predictable
The study also examined how business leaders view AI’s impact on employment.
The proportion of executives saying it was too early to determine AI’s effect on jobs increased from 4% in 2024 to 19% in 2026.
This indicates that as AI capabilities evolve, predicting the long-term impact on workforce requirements has become more difficult.
AI can automate repetitive activities, assist employees with complex tasks and create demand for new technical capabilities. However, the extent to which these developments will replace jobs, change existing roles or create new opportunities remains uncertain.
TCS also highlighted weaknesses in workforce planning. Only 28% of respondents reported maintaining an enterprise-wide framework or evolving skills taxonomy.
Such frameworks help businesses identify the skills employees need as AI changes job responsibilities and workflows.
Without structured workforce planning, companies may struggle to move employees into new roles or provide the training required to work effectively alongside AI systems.
What the TCS Survey Reveals About Enterprise AI Adoption
The study surveyed 1,333 senior business and technology leaders across 25 countries and 12 industries. The respondents included C-suite executives, divisional leaders, senior vice presidents and vice presidents.
Their organisations reported annual revenues ranging from $500 million to more than $20 billion.
| Key finding | Survey result |
|---|---|
| Executives identifying productivity as the leading AI value | 84% |
| Enterprises allowing multi-function autonomous AI decisions | 10% |
| Organisations with securely orchestrating AI agents | 9% |
| Respondents prioritising model independence | 90% |
| Organisations with AI provider-switching continuity plans | 7% |
| Enterprises assessing AI costs against value | 39% |
| Respondents reporting fragmented business knowledge | 44% |
| Enterprises deploying open-source AI selectively | 46% |
| Respondents identifying AI sovereignty as a major risk | 34% |
| Organisations with enterprise workforce skills frameworks | 28% |
The findings show that enterprise AI adoption is progressing, but the organisational foundations required for large-scale autonomous deployment remain uneven.
The Bigger Picture
The TCS study highlights an important transition in the global AI market. Businesses have moved beyond experimenting with chatbots and generative AI tools, but giving autonomous systems access to critical enterprise workflows requires a much higher level of trust.
The next competitive advantage may come from companies that can combine AI capabilities with secure data infrastructure, governance, flexible model architectures and effective workforce planning. Productivity improvements are an encouraging starting point, but enterprises will increasingly be judged on whether AI generates measurable financial and operational outcomes.
For IT services companies such as TCS, this transition also creates an opportunity to help clients redesign business processes, integrate AI into enterprise applications and establish the controls needed for autonomous operations.
Looking Ahead
The next phase of enterprise AI adoption will likely focus on moving agents from experimental environments into carefully controlled production workflows. Businesses will need to establish clear permission structures, improve access to internal information, monitor AI decisions and determine which activities require human approval. Progress in these areas could gradually increase the number of organisations comfortable with autonomous AI systems.
For the wider technology industry, the findings suggest that building more capable AI models is only part of the challenge. Companies must also solve problems involving security, cost, data integration and accountability before AI agents can operate independently at scale. The organisations that successfully address these issues could be better positioned to convert today’s productivity gains into sustained business growth.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
