As AI moves from assisting humans to making decisions, enterprises face new questions around accountability, governance and business value. In an email interaction Chanakya Bellam, Whole-Time Director, AION-Tech Solutions, shares his insights on the rise of agentic AI, India’s enterprise readiness and the importance of human oversight.
We have seen AI evolve from answering questions to performing tasks and increasingly making decisions. What does this shift mean for enterprises, and how prepared are Indian businesses for it?
The shift from AI that primarily answers questions to AI that can execute tasks and increasingly support decision-making represents a fundamental change in how enterprises use technology. AI is moving from being an assistive layer to becoming an active participant in business processes, with the potential to influence customer service, operations, risk management, forecasting and resource allocation.
For enterprises, this means moving beyond the deployment of AI tools to building the right data, technology and governance foundations. Agentic and decision-oriented AI systems require access to reliable, contextual and timely enterprise data. They also need clearly defined controls specifying which decisions AI can make independently, which require human oversight and how those decisions can be audited. As AI becomes more embedded in workflows, data quality, security, explainability and accountability will be as important as the underlying AI models.
Indian businesses are increasingly prepared for this transition, particularly large enterprises that have already invested in cloud, data platforms, automation and analytics. However, readiness remains uneven across industries and organisations. Many businesses continue to operate with fragmented data environments, legacy systems and siloed processes, limiting their ability to adopt more autonomous AI effectively.
The next phase of enterprise AI in India will depend less on adopting increasingly sophisticated models and more on strengthening the ecosystem around them. Organisations that combine high-quality data, modern technology infrastructure and responsible AI practices will be better positioned to achieve measurable business outcomes. The key will be determining where AI can act independently, where human oversight is necessary and how the two can work together effectively.
Agentic AI is being positioned as the next big transformation in enterprise technology. Beyond the industry buzz, where do you see genuine business value emerging?
Agentic AI will create genuine business value when it moves beyond being an interface for accessing information and becomes embedded in processes where it can take action within clearly defined boundaries. The opportunity is particularly significant in areas where enterprises handle high volumes of data, repetitive workflows and decisions that require continuous monitoring and response.
Across industries, AI agents could identify anomalies in financial transactions, support risk and compliance workflows, optimise supply chains, anticipate operational issues, assist customer-service teams and dynamically allocate resources. In sectors such as BFSI, manufacturing, logistics, retail and healthcare, value will increasingly come from connecting enterprise data with workflows to enable faster, more context-aware decisions.
The business case for agentic AI should be measured by outcomes rather than the number of agents deployed. Relevant indicators include reduced processing time, improved operational efficiency, faster decision cycles, better customer experiences and stronger risk management. Achieving these outcomes requires reliable data, integrated systems and clearly defined governance frameworks.
The industry is still at an early stage, so enterprises should be cautious about granting AI unrestricted autonomy. A practical approach is to begin with decision support and human-in-the-loop workflows, allowing greater autonomy as systems demonstrate reliability. Agentic AI is likely to deliver the most sustainable value when it complements human expertise rather than attempting to replace it outright.
Ultimately, the objective is not greater autonomy for its own sake, but more responsive, data-driven and efficient business processes, with AI acting at the appropriate stages and within clearly defined boundaries.
When an AI agent makes a decision that results in financial loss, operational disruption or a customer complaint, who should be held accountable? How should enterprises establish clear boundaries for autonomous decision-making?
Accountability cannot be delegated to an AI agent simply because a decision was made autonomously. When an AI-driven decision results in financial loss, operational disruption or a customer complaint, responsibility must ultimately remain with the enterprise and the people responsible for designing, deploying and governing the system.
Enterprises need to establish clear decision boundaries before deploying AI agents. Low-risk, reversible and routine decisions may be automated, while decisions involving significant financial exposure, regulatory compliance, customer rights or reputational impact should require appropriate levels of human oversight.
This requires an AI governance framework that defines system ownership, authorised decisions, data access permissions, thresholds for human approval and procedures for recording and reviewing decisions. Audit trails are particularly important for understanding not only what decision was made, but also the data and rules that influenced it.
A risk-based approach is essential, as different applications require different levels of autonomy. An AI agent managing routine operational workflows may be granted greater independence than one influencing lending, insurance, hiring or other high-impact decisions. Enterprises must also monitor these systems continuously after deployment, as model behaviour, data and business conditions can change over time.
Ultimately, autonomous AI should operate within an enterprise-defined ‘decision envelope’. The objective is to establish where AI can act independently, where it must seek approval and when it should stop and escalate. This combination of technological capability, human oversight and governance will be critical to building trust as agentic AI moves into mainstream enterprise operations.
As AI systems gain greater autonomy, concerns around data security, regulatory compliance and governance become more important. Are Indian enterprises equipped to manage these risks?
Indian enterprises are becoming increasingly aware of the risks associated with AI deployment, although their readiness remains uneven across industries and organisations. As AI systems become more autonomous, managing these risks becomes more complex because enterprises must oversee not only how data is processed, but also how systems interpret information, make recommendations and take action.
The immediate priority is to strengthen the foundations on which AI systems operate. This includes data governance, access controls, privacy and security mechanisms, model monitoring, and clear accountability for AI-driven decisions. Enterprises must also understand where their data is being used, which models and applications can access it, and whether those systems comply with regulatory requirements and internal policies.
Larger organisations, particularly in BFSI, healthcare and telecommunications, have made considerable progress because they already operate within relatively mature compliance and risk-management environments. However, the rapid adoption of generative and agentic AI is introducing new challenges, including shadow AI, uncontrolled use of enterprise data, model-related risks and uncertainty around accountability.
To address these challenges, enterprises need to move from reactive AI governance towards an approach that is built into the entire AI lifecycle. Before deployment, they should establish clear rules governing data access, permitted decisions, required human intervention and the circumstances in which a system must stop or escalate an action.
For agentic AI solutions, enterprises also need dedicated capabilities such as Agent Governance and MCP Gateway tools to ensure the secure, controlled and governed use of AI agents. These capabilities can help enforce access controls, govern interactions between AI agents and enterprise systems, and ensure that agentic solutions operate within defined security policies and governance frameworks.
Indian enterprises have the capability to manage these risks, but technology alone will not be sufficient. AI governance must be a shared responsibility across technology, cybersecurity, data, legal, compliance and business teams. Organisations that combine innovation with strong governance will be better positioned to scale AI responsibly and maintain stakeholder trust.
Could you share a real-world example of an enterprise successfully deploying agentic AI? What measurable business outcomes did it achieve, and what lessons can other organisations draw from it?
One instructive example of enterprise AI adoption is the use of AI agents in customer service and IT operations. These systems can interpret requests, retrieve relevant information, initiate workflows and resolve routine issues with limited human intervention. Such applications demonstrate how agentic AI can progress beyond generating responses to completing tasks within defined operational boundaries.
The business value of these deployments should be assessed using measurable indicators such as resolution time, cost per interaction, first-contact resolution rates, employee productivity and customer satisfaction. However, enterprises must distinguish between improvements resulting from conventional automation, generative AI and genuinely autonomous AI agents when evaluating outcomes.
The key lesson is to begin with a clearly defined business problem rather than adopting agentic AI simply because the technology is gaining momentum. Workflows involving repetitive tasks, well-defined rules and measurable outcomes are often suitable starting points. Reliable enterprise data, integration with existing systems and appropriate human oversight are also essential for sustainable results.
Successful deployment depends not only on the capabilities of an AI model, but also on how effectively it is integrated into business processes. Enterprises should establish baseline performance metrics, introduce AI agents in controlled environments and monitor their decisions before expanding their autonomy based on demonstrated reliability.
Ultimately, successful agentic AI adoption should demonstrate measurable improvements in business performance while maintaining security, accountability and customer trust. For Indian enterprises, an outcome-led approach will be critical to translating the potential of agentic AI into tangible business value.
Looking ahead, do you believe agentic AI will fundamentally change how enterprises operate, or will its adoption remain limited to specific business functions over the next few years?
Agentic AI is likely to significantly influence how enterprises operate, although adoption over the next few years will be gradual and initially concentrated in specific business functions. Rather than transforming entire organisations overnight, AI agents will increasingly become part of everyday workflows, helping businesses move from reactive processes towards more proactive, data-driven operations.
In the near term, adoption is likely to increase in customer service, IT operations, financial reporting, compliance monitoring, supply chain management and predictive maintenance. These functions offer clear opportunities to improve productivity, reduce processing time and respond to operational issues more quickly. As AI agents become more reliable and enterprise systems become better integrated, their role could expand across interconnected business processes.
However, the pace of adoption will depend on more than technological advancement. Enterprises must address the integration of legacy systems, cybersecurity, accountability and the reliability of their data. In regulated or business-critical environments, human oversight will remain essential, particularly when decisions carry significant financial, operational or reputational risks.
A more fundamental change will be how work is organised. Employees will increasingly move away from routine, repetitive tasks towards supervising AI-driven workflows, interpreting insights and handling decisions that require judgement and contextual understanding. This shift will also require organisations to rethink workforce skills, process ownership and performance measurement.
For Indian enterprises, agentic AI presents an opportunity to improve operational efficiency and expand capabilities without proportionately increasing complexity. The greatest benefits are likely to come to organisations that adopt the technology with clear business objectives, measure outcomes and expand autonomy in line with demonstrated reliability.
Ultimately, agentic AI may not replace existing enterprise operating models entirely, but it is likely to reshape them significantly. Its long-term impact will depend on how effectively businesses combine autonomous capabilities with human judgement, trusted data and responsible governance.
