ADVERTISEMENT
-B Swaminathan | IMAWS
Artificial intelligence has firmly entered the boardrooms of Indian enterprises. Yet, for many Chief Information Officers, the distance between a promising AI demo and a working production system remains frustratingly wide. At the sidelines of a recent technology meet in Pune, four senior technology leaders spoke candidly about what it really takes to make AI work inside an organisation — and why the journey is harder than it looks. From ROI pressures and data quality concerns to governance challenges and change management, the conversations revealed that AI adoption in Indian enterprises is progressing, but still needs the right foundations to deliver real value.
PUNE: A MATURING TECHNOLOGY HUB
A common theme across all four conversations was the distinct character of Pune as a technology market. Leaders noted that organisations here prefer a structured, process-driven approach to technology decisions — spending time to carefully define requirements, evaluate multiple vendors in parallel, and go through formal RFP processes before committing. This stands in contrast to faster-moving markets where companies rely on a fixed set of trusted partners and move quickly.
This approach reflects Pune’s evolution from an automobile and education hub into one of India’s most active IT and Global Capability Centre (GCC) destinations. With a growing number of multinational companies setting up operations here, governance, data compliance, and adherence to both Indian and international standards have become just as important as technological capability. The result is a city where technology leaders must balance innovation with rigorous accountability.
The Demo-to-Production Gap: Why AI Keeps Falling Short
The reason is simple: demos are built on clean, carefully prepared sample data, so the AI performs well. But in production, data is messy, inconsistent, and complex. The system that worked smoothly in the demo begins to produce unreliable results, and the effort required to fix those gaps drives up costs sharply. When costs rise without a corresponding improvement in output, the business case weakens quickly — and what looked like a smart investment starts to feel hard to justify.
Phadke notes that this is a shared challenge across the industry. The gap between a successful demonstration and a scalable, cost-effective production system is one of the most common sticking points in AI adoption today. On the procurement side, however, he sees fewer concerns. Most organisations have already developed mature and well-defined procurement processes — some centralised, some delegated — and these work well. It is AI implementation itself, he says, that remains genuinely new and challenging territory for everyone.
*ROI First, Then Everything Else*
Paramanand Shinde, Vice President of IT at Sanghvi Movers , takes a clear-eyed and practical approach to technology investment. Every rupee spent must come with a demonstrable return. Without a well-built ROI case, he says, management will not approve the budget — and rightly so. His process is straightforward: establish the business case, secure the budget, and only then move to procurement and implementation.
The second challenge he highlights is change management. Technology implementations often succeed technically but fail in practice because people do not change how they work. Getting employees to adopt new systems requires active involvement from senior leadership — and that support is not always easy to secure. When leaders are not visibly behind a change, adoption at the ground level suffers.
Shinde also observes a common problem among his peers: organisations where business processes remain manual and disconnected. Data sits in spreadsheets, decisions get delayed, and teams escalate to IT only when they need something quickly — rather than having integrated systems that provide real-time insight. He also points out that a smart approach to AI readiness is to first understand how employees are already using freely available AI tools in their day-to-day work. Studying those usage patterns across functions like finance, HR, and marketing can provide a strong, data-driven foundation for making the case to management for a more structured AI investment.
Governance, Compliance, and Getting the Data Right
Parag Deshmane, Manager of Core IT and Infrastructure Operations at Alfa Laval, sees a landscape of growing complexity for CIOs in Pune. As the city attracts more multinational companies and GCCs, technology leaders must navigate a wide range of vendor choices, governance requirements, and compliance obligations — all while trying to build a coherent AI strategy that makes sense to the board.
One of the more pressing challenges is the regulatory environment. Multinational companies operating in Pune must comply with both Indian laws and the standards of the countries they originate from. India is developing its own Digital Personal Data Protection Act (DPDPA), which operates separately from frameworks such as Europe’s GDPR. Navigating multiple regulatory regimes simultaneously — alongside varying taxation laws across different countries — creates a significant and ongoing compliance burden. Aligning AI adoption with all of these requirements, rather than treating it as a standalone technology initiative, is where much of the difficulty lies.
On whether AI can help address these complex challenges, Deshmane is realistic. AI can certainly play a useful role, but only if the data going into it is clean and reliable. He is direct: garbage in, garbage out. If an organisation feeds messy or inaccurate data into an AI system, it will get poor results out — regardless of how advanced the technology is. The real priority, therefore, is data quality. Organisations must ensure their data is accurate and well-governed before expecting AI to deliver meaningful outcomes.
Let Innovation Happen — But Put Guardrails Around It
Karishma Kumar, CIO of IT at Skoda Auto Volkswagen India, offers one of the most balanced perspectives on AI adoption. She begins with an observation about Pune’s technology culture: unlike faster-moving markets where businesses rely on established partners, Pune organisations take a more thorough approach — evaluating multiple vendors, going through formal RFP processes, and taking time to get requirements right. In the context of AI, this careful approach may actually work in their favour.
Across the industry, Kumar notes, a large share of AI use cases are still not succeeding. This makes the choice of use case critical. The right approach, she believes, is to identify pilots that can show results quickly and move to production faster, while continuing to support broader experimentation. She is firm that AI will not replace IT teams — it will enhance them. A task that once required four people and a week can be done by one person in an hour, but only when data and systems are properly connected. Without integrated systems as a foundation, she warns, AI cannot be expected to deliver.
What stands out in her approach is how she handles the balance between enterprise governance and ground-level AI innovation. Across industries today, business departments are running their own AI pilots independently — without always looping in the IT team. Rather than stopping these experiments, the right approach, she argues, is to introduce clear guardrails: let teams pilot freely, but ensure that before any tool goes live in production, it is reviewed and deployed on a secure, enterprise-grade platform. The goal is to keep innovation alive while protecting data security and maintaining accountability. Let the business explore, she says — but when it is time to go live, do it properly.