Discipline, governance and long-term strategy underpin the scaling of enterprise AI
At a recent roundtable in Chennai hosted as a part of AI@Work series organised byThe Hindu Group in association with IBM, technology, finance and healthcare leaders gathered to examine one of the most pressing questions facing enterprises today: how to translate AI investments into measurable, sustainable business value.
The discussion, moderated by Mr Suresh Vijayaraghavan, CTO, The Hindu, themed “Navigating AI investment and scalability: Strategies to drive ROI in an evolving landscape”, moved beyond the hype around artificial intelligence and into the hard realities of execution, risk, and long-term transformation. The conversation was structured around three pillars: the ROI imperative, scaling smart, and strategic foresight.
From pilots and hype to measurable returns in AI
The first section focused on a problem that almost every organisation now faces: AI experiments are easy, but sustained value is rare.
Sujatha S,Head of AI (Security) at Zoho, captured this gap succinctly. While proofs of concept are often built in ideal conditions with clean, well-structured data, the real world is far messier. “When you do a POC, you assume such picture-perfect conditions, you assume that you have perfect data, clean data. But when you take it to production, you have a plethora of challenges,” she said. When AI systems move into production environments, they must deal with inconsistent, incomplete and constantly changing data. This is where many AI initiatives struggle, not because the technology is weak, but because organisations have not built the operational and data foundations needed to support it at scale.For many firms, she noted, projects fail not because the models are weak, but because the data, workflows and governance are not production-ready.
From a financial lens, Rohit Kumar Agrawala, Director – Finance at Chennai Petroleum Corporation Limited, reframed how executives should evaluate AI bets. “Unless you understand the business, you can’t understand what kind of framework to use, what kind of value it emerges and what kind of applications or tools to use” he said, emphasising that AI must be judged against real financial thresholds, not just technological novelty. His point was that AI should not be pursued simply because it is fashionable; it must generate returns that clearly exceed financial thresholds, otherwise it becomes a poor business decision regardless of how advanced the technology may be. In other words, successful AI investment begins not with technology, but with clarity about business priorities, processes and outcomes.
Ajay Thomas John,Group CDO at the Shriram Group, argued that ROI must be linked to business maturity. “For us, our success curve is much more important than anything else. We define technology in terms of which part of your success curve is this technology on,” he said. Not all AI investments are ready for full-scale deployment; some belong in experimentation, others in core operations.He emphasised on the fact that not every AI investment needs to be enterprise-wide immediately; some technologies belong in early-stage pilots while others are ready for mainstream deployment. Understanding where a technology sits on this curve helps leaders make smarter decisions about funding, risk and expectations.
In healthcare, the case for ROI was far more immediate. Dr Ilankumaran Kaliamoorthy, CEO – Chennai Region, Apollo Hospitals, gave a striking example: “We have started using AI-assisted reporting of X-rays. Previously, it used to take at least four hours… now the time has come down to less than 20 minutes.” Here, ROI was not just financial, but measured in patient experience, clinical throughput and quality of care.This not only improves patient experience but also increases hospital efficiency and clinical productivity. His example demonstrated that in sectors like healthcare, ROI from AI is measured not just in money, but in time saved, lives improved and quality of care delivered.
Smart scaling of artificial intelligence through culture and governance
If ROI is about proving value, scaling is about protecting and multiplying it.
Santhosh T G,Global CDO at Switch Mobility, said that organisations must think beyond quick wins. “For a long-term perspective, it’s only two things we think of — one is to build culture over a period of time and two is build capabilities,” he said. In fast-moving AI environments, sustainable advantage comes from people and processes as much as from algorithms.He argued that organisations must consciously invest in building an innovation culture and strong internal capabilities over time. Without these foundations, even the best AI solutions will struggle to deliver value, because there will be no organisational muscle to absorb, adapt and scale them sustainably.
For Durgaprasad Swaminathan, EVP and CIO at Cholamandalam Investment and Finance, organisations fall into distinct roles. “You fall into three categories — either you are a taker, a shaper, or a maker, or a combination of these,” he said. Firms must consciously decide whether they are merely consuming AI tools, shaping them to their needs, or actively building differentiated platforms.Some companies simply consume ready-made solutions, others adapt and shape them, while a few build their own platforms. Each model carries different levels of cost, control and competitive advantage. The key is to make this choice deliberately, rather than drifting into it without a clear strategy.
Operationalising AI also requires new forms of accountability. Rohit Sood, Advisory Data & AI Specialist at IBM, warned against over-automation without safeguards. “If you’re bringing in agentic AI to bring decision-making into your process, it’s actually a human-in-the-loop responsibility which needs to be ensured — otherwise you will never bring it into production.”Without a “human in the loop,” organisations risk losing control, accountability and trust. This governance is essential if AI systems are to be safely deployed in real business environments.
For banks, regulation and trust remain paramount. Sankaran G, CIO at City Union Bank, said, “Before launching any product, we need three segregations — one for regulatory, one for competency in the market, and third for customer convenience.” AI, he argued, must fit cleanly within this governance framework.Even the most advanced solution cannot be launched unless it satisfies all three. This structured approach helps banks manage risk while still innovating.
Satyen Kumar Jadeja, Consulting Strategic Sales Lead BFSI, IBM shared a real-world example of what smart scaling can achieve. “With the help of AI, we cut one bank’s processing time from 28 hours to 15 minutes — their capacity is almost five times now,” he said, illustrating how operational gains can translate directly into business advantage.By automating and optimising workflows, IBM helped a bank reduce processing time from 28 hours to 15 minutes, effectively multiplying its throughput without adding manpower. This demonstrated how well-implemented AI directly translates into business scalability and efficiency.
Strategic foresight on talent architecture and long-term risk
The final section turned to the future, and to the risks of both action and inaction.
Jegadeeswaran B, Senior General Manager – IT at TVS Automobile Solutions, highlighted the importance of blended skill models. “We train the internal talent pool along with the partners so that the skill coexists with the internal talent,” he said, describing a hybrid approach to capability building.While external partners bring expertise, companies must also train their own employees so that skills stay within the organisation. This shared model ensures that AI knowledge becomes part of the company’s DNA, allowing it to innovate faster and avoid long-term dependency on vendors.
Nagaraj Nagabushanam, Vice President – Data and Analytics and Designated AI Officer, The Hindupointed out that GenAI is uniquely accessible and that it has lowered the barrier to entry because people can now interact with technology using natural language rather than code. “One of the major advantages, especially in the GenAI space, is that English is your programming language,” he said. But he also warned of deeper architectural risks. “Technical debt, I’m keenly aware that the entire architecture is changing… what we took for granted earlier may no longer hold.”Organisations must therefore be careful not to accumulate technical debt by building on fragile or outdated systems that may soon become obsolete.
In healthcare IT, Ravi Balakrishnamurti, CIO at Sundaram Medical Foundation, said his organisation wanted to be proactive. “We would like to start as a maker itself — to be exposed to certain tools,” he said, rather than waiting passively for off-the-shelf solutions.By exposing themselves to AI tools and building internal understanding, hospitals can design systems that are better aligned with their workflows, data and patient needs rather than relying entirely on generic solutions.
Dr Sindhuja G, Digital Lead at Sundaram Medical Foundation, added that AI adoption must begin with clinicians’ pain points. “Data entry itself used to be a pain point for us,” she said, underlining that automation must reduce burden, not add to it.Doctors spend excessive time entering data instead of caring for patients. AI can relieve this load by automating documentation, allowing clinicians to focus on what truly matters: patient care.
For Sudheer Warrier, CIO at Sundaram Finance, discipline was the key to future-proofing. “Many times, we create debt because we are trying to do something with AI, not trying to solve the problem,” he said. Poorly designed AI, he warned, creates systems “that are very difficult to change later.”When organisations adopt AI without clearly defining the problem they are trying to solve, they end up with complex systems that are hard to change or upgrade. Thoughtful design at the start is critical to avoid future constraints.
Moderating the discussion, Suresh Vijayaraghavan, CTO, The Hindu, brought the conversation full circle by introducing the idea of risk of not investing. “When we don’t invest for a period of time, we are accumulating technical debt and financial debt by not investing,” he said — a reminder that standing still can be just as dangerous as moving too fast.By delaying AI adoption, organisations accumulate both financial and technical debt, making it harder to compete in the future. His message was that leaders must balance caution with urgency, waiting too long can be just as costly as moving too fast.
From pilots to purpose
Across industries like banking, healthcare, mobility and media, the panel made it clear that the true return on AI does not come from how many models an organisation deploys, but from how precisely it applies them to real business problems. AI creates value only when it is directed at the right use cases, implemented at the appropriate scale, and governed with discipline. Without this focus, even the most advanced technologies risk becoming expensive experiments that fail to move the needle on productivity, customer experience, or profitability.
As organisations progress from small pilots to enterprise-wide adoption, the centre of gravity shifts from technology to execution. Data readiness determines whether AI canperform reliably, governance ensures that decisions remain compliant and accountable, talent provides the capability to build and adapt systems, culture shapes how widely and confidently teams adopt them, and strategic clarity keeps investments aligned with long-term business goals. These elements form the foundation that allows AI to mature from a promising tool into a durable competitive advantage, one that continuously improves operations, enhances decision-making, and enables organisations to stay ahead in an increasingly digital and data-driven world.
