80% of Indian organisations are already exploring agentic AI, autonomous agents capable of pursuing goals independently.
Just as you’re preparing to sign off for the weekend, an email from the chief executive lands in your inbox. The subject line is innocuous, but the message is anything but: “What’s our plan for agentic AI?” With that, your evening changes course.
Buzzwords—“autonomous agents,” “multi-step workflows,” “generative AI”—begin to swirl in your mind. You recall tales of agentic AI systems navigating complex processes with minimal human input. Some ended in impressive efficiency gains; others fizzled amid integration chaos and ballooning costs. Is your CEO merely curious, or is this the start of yet another urgent technology mandate?
The question is not unusual. According to Deloitte’s latest State of GenAI report, more than 80% of Indian organisations are already exploring agentic AI—autonomous agents capable of pursuing goals independently. Over half identify multi-agent orchestration as a strategic priority, and 70% plan to harness GenAI to automate workflows. In a sign of the times, 71% are juggling ten or more generative AI experiments simultaneously.
This is the new reality for enterprise IT leaders: just months after grappling with the implications of GenAI tools that compose emails or summarise documents: They now face its more operationally ambitious cousin. Agentic AI promises more than content creation. It pledges intelligent action.
As Mukesh Rathi, Global CIO of Hinduja Group, puts it: “The moment to act is now… Agentic AI has reached a level of maturity where early movers can gain a clear competitive edge.”
This month’s cover story delves into the emerging paradigm of Agentic AI—how it builds on, yet diverges from, traditional generative AI, and what it signals for the next wave of enterprise automation. We also ask the critical question: Are CIOs prepared to lead this transformation, or will they risk falling behind in a rapidly evolving digital landscape?”
Why Agentic AI matters
Agentic AI, also referred to as autonomous AI agents, are software solutions capable of completing complex tasks and achieving defined business objectives with little to no human intervention. Agentic AI differs significantly from today’s chatbots and co-pilots.
Think of generative AI as the creative team in your organization, and agentic AI as the operational crew. You can assign an agentic AI a goal—such as responding to a customer inquiry, adjusting a logistics schedule, or sending follow-up emails—and it will take the necessary steps to accomplish that objective.
However, greater autonomy introduces greater complexity. Agentic AI systems can chain together multiple steps, dynamically shifting course as new information becomes available. They’re not limited to executing a single, narrow task—they’re orchestrating entire workflows.
“The playbook doesn’t change—whether it’s RPA, machine learning, traditional AI, or agentic AI,” Rathi adds. “Start by selecting the right use cases—those with clear business relevance and measurable impact. The advantage today is the breadth of ready‑made agentic AI offerings. You don’t need an extended R&D cycle to prove feasibility.”
The quiet shift from curiosity to adoption
The industry is undergoing a significant evolution—from single-step automation tools like chatbots to sophisticated, multi-step, agentic systems. While a chatbot might handle basic queries, an agentic system can execute complex workflows: retrieving data from multiple sources, generating reports, distributing them, and even following up on pending actions.
“Wherever there’s a repetitive task and sufficient data, AI agents can absolutely be applied in manufacturing,” says Anand Deodhar, Group CIDO at Force Motors. “Start with chatbots. Once you’re comfortable, scale across different functions. But the key is quality data—if your data is poor, the AI’s output will be, too.”
Examples abound across sectors. Banks are automating intricate processes like guarantee management. Manufacturers are using AI for tasks such as invoice matching. In healthcare, multi-step agents are assisting with image analysis and prescriptions. Automotive players are streamlining the lead-to-customer journey. And these use cases are just scratching the surface.
Apoorv Iyer, Executive Vice President and GenAI Practice Head, HCLTech, calls Agentic AI a game-changer. He emphasizes that it’s not just about cost reduction or automation—it’s about enhancing productivity so skilled employees can focus on strategic work. HCLTech is actively co-innovating with clients to unlock AI’s potential across the enterprise.
One of their standout initiatives is the development of reusable, modular AI agents. These include orchestration agents for managing data flows, modernization agents for updating legacy systems, and agents for specific use cases like retail return management. Designed for flexibility, these agents can be seamlessly embedded into various parts of a business process.
Iyer advises CIOs to focus on three key areas to get ready for agentic AI:
Strategy – Is your AI initiative aligned with business goals?
Infrastructure – Is your tech stack capable of supporting AI workloads?
Workforce – Do your teams have the skills to thrive in an AI-augmented environment?
Another compelling example comes from Kogo, where Raj K Gopalakrishnan and his team are developing an “agentic mesh”—a network of interoperable AI agents that share memory and context using a universal communication protocol. This is in stark contrast to traditional siloed AI assistants that handle only narrow tasks.
At the recent AutoExpo, Kogo demonstrated this concept by integrating MapMyIndia’s geointelligence into an in-car assistant. Ask for directions and the navigation agent kicks in. Then switch topics to a sushi recipe, and a food agent takes over. Ask about sports scores, and it transitions again—without breaking the flow. You can even check emails or place a grocery order. There’s no need to switch between apps—the agents talk to each other, passing context along the way.
The big breakthrough? These specialized AI agents not only collaborate but also remember what the user asked moments ago. Mention the weather, and the system understands you’re still on the sushi thread from a minute earlier. This contextual continuity is what makes agentic AI transformative. It’s not just smarter automation—it’s a mesh of connected intelligence, working as one.
The security angle
Early in your research on agentic AI, you’ll likely ask, “Is my data safe?” Security is a top concern with large language models (LLMs) and complex workflows.
Raj K Gopalakrishnan from Kogo addresses this by highlighting that their system can run on-premise. Personal data stays in your private infrastructure when the agentic mesh is deployed. He warns that sending every data point to a public LLM means effectively standing naked in front of the world.” That’s not something big enterprises, or their regulators, are comfortable with.
HCLTech offers a similar outlook. They see a layered approach, blending a secure cloud environment for specific tasks with on-site controls for sensitive data. By controlling how data flows between AI agents, they reduce the risk of exposure.
Data privacy is non-negotiable for industries like banking, defense, and healthcare. The good news is that agentic AI can be deployed in ways that respect those requirements. The challenge, of course, is designing your stack and governance structures up front. That means working closely with legal, compliance, and data governance teams from the start.
“AI is going to fail without a strong data foundation,” cautions Jay Upchurch, Executive Vice President and CIO at SAS. “Modernize your data estate, put governance in place, and drive AI literacy across the company. That’s probably the number‑one reason why AI aspirations fail today.”
Barriers to scaling
Despite the excitement, it’s essential to acknowledge the scaling challenges. Deloitte found that only 29% of organizations can scale even 30% of their AI proofs of concept. That’s shockingly low if you consider how many AI pilot projects are out there. Why do so many stalls?
- Tokenization costs: Large language models often charge based on the number of words or tokens processed. With agentic AI, your system might make many queries, and costs can pile up quickly.
- Integration roadblocks: Agentic AI isn’t just a single app. It’s a web of connections to CRMs, ERPs, data lakes, and more. Older systems may not have easy APIs, and the engineering effort can be huge.
- Data quality: Agentic AI thrives on reliable data. The AI’s decisions will be suspect if your source systems are messy. Anand agrees, “What’s crucial is ensuring high‑quality data; if your data is poor, the AI outputs will also be poor.”
- Regulatory constraints: In regulated spaces, every step in a process might need logging. Agentic AI can take varied paths to solve a problem, making it harder to track. Regulators might require an “explainable AI” approach that agentic systems aren’t always ready to provide.
- Talent gaps: Middle managers often lack the skills to guide AI projects. Senior leadership might be enthusiastic, and junior data scientists might love the technology. Still, the team in the middle—the ones who must operationalize it—may not know how to integrate it into day-to-day workflows.
“The AI journey for us is starting as of now,” notes L B Sharma, CIO at Bharat Petroleum Corporation Limited. “We are still trying to figure out data consolidation and quality assessment. Whatever pilot projects we’ve conceptualized are being validated, so we don’t end up in violation. It remains in the pilot for now, but we plan to go to production soon.”
The four supplier “Houses” leading the market
Ashvin Vellody, Chief Strategy Officer at Deloitte, categorizes agentic AI providers into four distinct groups, which he refers to as “houses” (borrowing from Game of Thrones):
- The first house consists of OEM vendors. For enterprise IT leaders, existing software providers are among the most likely avenues through which agentic AI will enter organizations. These include ERP, CRM, product, API, and process automation vendors. Companies like SAP, Oracle, Microsoft, Salesforce, Adobe, and ServiceNow are all embedding agentic capabilities within their platforms.
- The second house includes the hyperscalers—the “kings of the cloud.” These providers offer tools and capabilities through AI-powered Platform-as-a-Service (PaaS) environments. Notable examples include Amazon’s Bedrock, Google’s Agent Space, Microsoft’s Copilot ecosystem, and similar offerings from other cloud providers.
- The third house is open source. While offering unmatched flexibility and control, open-source approaches typically entail higher development costs and demand specialized talent. Organizations adopting this route can build their own AI stack, but success requires significant expertise.
- The fourth house comprises startups focused on vertical specialization. These players create agents tailored to specific industries or use cases. While they promise faster time-to-value, they also carry the risk of business failure or acquisition, which could disrupt service continuity.
Each of these approaches offers a unique strategy for implementing agentic AI, with different trade-offs around integration complexity, customization flexibility, and vendor lock-in.
The fundamental conflict at the heart of adoption
Ashvin Vellody points to a central dilemma that every CIO must address: technology is evolving at breakneck speed—every six months, compared to Moore’s Law’s 18–24-month cadence—while organizational change remains slow and deliberate.
The pace of AI innovation far exceeds what most enterprise IT teams are structured to handle. Regardless of the LLM stack, the GPUs chosen, or the hyperscaler engaged, new capabilities are emerging so quickly that continual investment is required just to stay current.
Meanwhile, organizational change management remains stubbornly difficult. As Vellody noted in his presentation:
“Organizations change very slowly. There is no way organizations change fast… No matter what you say or how many conferences we attend, the change in technology in organizations is slow.”
This creates a genuine dilemma. Wait too long, and risk falling behind competitors who move faster. Dive in prematurely, and face the possibility of expensive failures that set your AI program back by years.
The financial stakes are considerable. Deloitte finds that 94% of firms would need more than six months to exit a failed AI initiative, and 76% estimate it would take over a year. Meanwhile, 28% of organizations fear their current AI solutions will be obsolete within two years, making AI adoption both urgent and risky.
Are CIOs missing out if they wait?
So, should CIOs prioritize agentic AI now, or is it safe to wait? The answer likely lies in a balanced approach.
There’s a cost to waiting. Early adopters gain valuable process insights and institutional learning. By the time others catch up, early movers may already have captured market share or built superior customer experiences. This is especially critical if your competitors are known to innovate rapidly.
At the same time, rushing in unprepared can be costly. Poorly scaled AI initiatives can drain budgets, particularly when vendors or models are misaligned with enterprise needs. Additionally, workforce readiness—especially in terms of cultural adaptation—can be a major bottleneck.
Humans + AI = A new work paradigm
A key theme in the agentic AI conversation is the human role. While many fear AI will replace jobs, most experts envision a partnership. Agentic AI handles routine tasks, data aggregation, and administrative workflows. Humans continue to drive strategy, creativity, innovation, and relationship-building.
Apoorv Iyer of HCLTech notes that AI is “transforming how we work, innovate, and grow.” By automating repetitive tasks, CIOs can refocus talent on strategic objectives. This doesn’t eliminate jobs—it reshapes them. For example, “data entry specialists” might evolve into “AI oversight managers,” or “process operators” into “process innovators.”
At Kogo, the concept is applied on the road: turning a car into a personal assistant. The driver retains control, but AI handles tasks safely, reducing friction and accelerating daily decision-making.
Manas Mehra, Global CIO at Dabur India Limited, underscores the need for thoughtful implementation:
“Agentic AI is something we cannot ignore—but there’s a journey to it. Models like ChatGPT, Copilot, or Gemini are highly trained. Still, once they enter your organizational ecosystem, you have to understand that what they offer is cognitive ability—and that takes time to build. As CIOs, part of our role is expectation management—clearly defining what will work in our organization, building the base, setting realistic expectations, and executing.”
Final thoughts
Agentic AI isn’t just a buzzword. It builds on generative AI foundations but ventures further into real-time decision-making and automation. CIOs globally are paying attention. The critical question isn’t if agentic AI will matter—it’s how soon.
Deloitte’s research indicates the shift is already underway. Companies like HCLTech and Kogo are seeing tangible benefits. In sectors from banking to healthcare to automotive, the use cases are multiplying—automated loan processing, patient data workflows, and advanced in-car assistants are just the beginning.
Challenges are real—security, data quality, regulation, and scaling all demand attention. But those who prepare properly can unlock unprecedented gains in productivity and innovation.
So ask yourself: Will you shape agentic AI for your enterprise—or let it shape you?
Now may be the perfect moment for CIOs to explore a pilot, align it with strategic goals, and build a governance framework. You don’t need to go all in—but ignoring it could mean missing out on competitive advantage.
As Mukesh aptly puts it, “The moment to act is now—early movers on agentic AI will lock in a clear competitive edge.”
With adoption projected to grow at a 43.8% CAGR through 2034, agentic AI isn’t going away. The technology—and the competition—are both moving fast. The conversation has already begun. Whether you leap now or wait, your next email from the CEO may just ask:
“What’s your AI plan?”