Agentic AI Is Key To Unlocking Real, Scalable Impact Across Enterprises, Says IBM’s Siddhesh Naik

Agentic AI is one of the top strategic technology trends today. Gartner predicts that by 2029, agentic AI will autonomously resolve 80% of common customer service issues without human intervention, leading to a 30% reduction in operational costs. We spoke to Siddhesh Naik, Country Leader, Data & AI Software, IBM India & South Asia, on how enterprises are leveraging agentic AI, what IBM is doing in the space, the security considerations CIOs should be wary of as they deploy AI agents, and the future of agentic AI.

Q. What is the role of agentic architecture in the evolution of AI? How is it different from traditional AI like chatbots?

Agentic architecture is moving beyond task-specific automation to truly goal-driven systems. Traditional AI, like chatbots, are mostly reactive. They respond to inputs within a limited context and cannot plan or adapt. Agentic AI, on the other hand, is designed to operate with intent. It can plan, act, reflect, and improve over time, much like how a human would approach a complex problem.

At IBM, we believe this shift is critical for enterprises that are looking to scale AI responsibly. Agentic architectures power AI agents that can work across functions, call tools, access real-time data, and even collaborate with other agents to solve complex tasks, all with minimal human input. This is what we are enabling through our enterprise-ready AI and data stack, watsonx platform, that brings trust, transparency, and governance to the way agentic AI is deployed. 

Q. Can you give some examples of agentic AI at work today in customer-facing businesses, and some ideas about the possibilities?

Agentic AI is already delivering tangible value across front end use cases. In retail, AI agents are powering hyper-personalised product recommendations, managing dynamic pricing, and autonomously optimising inventory across channels, leading to better conversions and fewer stockouts. In banking, AI agents monitor real-time risk, support continuous compliance checks, and enable personalised wealth management by analysing customer portfolios and market signals without human input.

In sales, agentic AI qualifies leads, automates follow-ups, and supports reps by surfacing key insights during live interactions, improving win rates and shortening deal cycles. These agents are no longer just back-end tools; they are becoming front-line collaborators, enabling businesses to respond faster, personalise at scale, and reduce operational friction. 

Q. How does agentic AI benefit consumers and businesses? What are the pros and cons?

Agentic AI represents a major leap in how businesses use automation and intelligence. Unlike traditional AI tools, which are often limited to handling one task at a time, agentic AI systems can plan, reason, and execute complex tasks on their own with minimal human involvement. This ability, to autonomously call on internal and external tools and collaborate with other technologies, has the potential to unlock a new level of efficiency and productivity for businesses. For consumers, the impact is equally significant. They benefit from more personalised experiences, whether it’s receiving timely product recommendations, proactive customer support, or AI-driven financial advice that adapts to their needs in real time.

For businesses, the benefits are clear — improved productivity, cost savings, faster decision-making, and higher accuracy. For example, we worked with Dun & Bradstreet to use our AI assistant capabilities from watsonx Orchestrate to develop D&B Ask Procurement. The assistant provides a conversational chat experience that automates repetitive tasks and simplifies complex procurement processes.

It gave them the capability to query deep data on suppliers and use advanced AI technology to interpret that data, while eliminating the need for time-consuming, manual data entry, thus reducing errors. Furthermore, watsonx Orchestrate provided the ability to integrate advanced automation capabilities to streamline work across systems and enhances the overall efficiency and effectiveness. 

Q. How is IBM leveraging agentic AI to gain a strategic advantage? Any IBM deployments in India you can share about?

At IBM, we believe agentic AI marks a fundamental shift, from AI that simply chats or assists, to AI that can truly take action. We are focused on building what we call the agentic enterprise, where AI agents are embedded across the entire technology landscape to help users get work done faster, smarter, and with greater autonomy. With IBM watsonx Orchestrate, we are enabling organisations to put AI agents to work across existing systems and tools, from Salesforce and SAP to Workday and ServiceNow. What sets us apart is our ability to integrate seamlessly across the entire Application Landscape, across OnPremise and multiple Clouds and help businesses get value from both new AI investments and their existing IT estate.

Q. What are the security considerations business and technology leaders should factor when deploying/building agentic AI solutions. What are the challenges when it comes to agentic AI vis-à-vis ethical AI?

As agentic AI systems become more autonomous, making decisions, taking actions, and learning on the fly, the risks around security and ethics grow significantly. From a security perspective, agentic systems often rely on APIs to interact with external data and applications. These interfaces must be secured with robust access controls and authentication mechanisms to prevent unauthorised access or data leaks. The risk of adversarial attacks, where manipulated inputs mislead the AI, also increases, especially in high-stakes environments like finance or healthcare. Maintaining a strong security posture is crucial, from securing Active AI deployments to identifying Shadow AI within the organisation.

On the ethical front, agentic AI brings new complexity. Unlike rule-based systems, agents make decisions through opaque machine learning processes, often without human oversight. This makes it difficult to ensure transparency, explainability, and fairness. For example, if a customer service agent adapts to negative interactions, it could begin responding with unintended tone or bias. Similarly, if an agent prioritises efficiency, it might overlook fairness or privacy considerations. 

IBM is addressing these challenges: (a) by enabling AI sandboxing to simulate agent behavior in controlled environments before real-world deployment, (b) introducing AI systems called governance agents, that oversee and moderate other agents in production, (c)stress testing agents under edge-case scenarios to identify ethical blind spots, and (d) incorporating specialised metrics like context relevance, answer similarity, and faithfulness into watsonx.governance to ensure responsible performance. 

Q. What does the future of AI look like with the rise of agentic systems?

We believe that agentic systems are key to unlocking real, scalable impact across enterprises in the near future. Hence, we are building the foundation for this future, combining generative AI with automation, real-time data integration, and enterprise-grade governance. Our focus is not just on what agents can do, but how they do it, ethically, securely, and transparently. It shifts the narrative from one-off use cases to continuous, outcome-driven automation, designed for scale, efficiency, and long-term value. This is where AI starts becoming truly useful in the enterprise, context-aware, always learning, and capable of driving business outcomes.