Sumit Duttagupta, CIO of Haldia Petrochemicals, shares insights on driving AI-led innovation through collaboration, inclusivity, and measurable ROI.
As AI moves from proof-of-concept to large-scale deployment, CIOs are dealing with a multifaceted challenge: integrating advanced technologies while managing workforce transitions, budget constraints, and operational efficiency. AI’s potential extends beyond automation—it is reshaping industries through predictive analytics, digital twins, and real-time decision-making. However, its true impact lies in how seamlessly organizations embed AI within a people-centric framework.
In an exclusive interaction with CIO&Leader, Sumit Duttagupta, CIO of Haldia Petrochemicals, shares his insights on driving AI-led innovation while fostering collaboration, inclusivity, and measurable ROI. From tackling legacy system integration to leveraging AI for demand forecasting and plant reliability, he outlines the strategic roadmap for AI implementation in 2025.
Sumit is a seasoned technology leader with a wealth of experience in senior leadership roles, including Group CIO, VP, and Head of IT. In his current role, he leads Haldia Petrochemicals’ digital transformation initiatives, emphasizing process optimization, automation, and the seamless integration of plant and business data. His efforts are geared towards enabling proactive, predictive, and prescriptive decision-making. Sumit is also a recipient of the prestigious 6th CIO&Leader Samman. Below are excerpts from the interview:
CIO, Haldia Petrochemicals
CIO&Leader: How do you see AI adoption evolving particularly in terms of balancing technological advancements with workforce transition and maintaining a collaborative work environment?
Sumit Duttagupta: 2024 was essentially a “wait and watch” year. Several key developments emerged, such as the maturity of AI as a technology. In our business, we also faced a highly volatile commodity market. Additionally, in the chemical and petrochemical industries, concerns about environmental sustainability deepened, prompting organizations to adopt ESG initiatives. While necessary, these efforts required significant investments in process re-engineering and long-term planning, as they are expected to impact business operations over the next 5 to 10 years.
Furthermore, escalating geopolitical tensions disrupted supply chains, driving up transportation and shipping costs.
Looking ahead to 2025, as these developments mature, this year is expected to be both challenging and decisive for new technology adoption. We anticipate that AI will have a profound impact on an organization’s people-centric approach, necessitating the development of integrated workplace architecture. At the same time, it is crucial to foster a workplace culture that embraces new technologies while promoting a data-driven analytics approach.
However, it is equally important to preserve the essence of a cohesive and collaborative work environment. The introduction of new technologies may empower certain “smart users” to work independently with confidence, but technology adoption should not lead to silos. Instead, it should enhance teamwork and collaboration within the organization.
CIO&Leader: How are technologies like digital twins, AR, and VR helping drive innovation and new business models?
Sumit Duttagupta: When it comes to adopting predictive technologies, we’ve already developed a digital twin for our naphtha cracker unit. This real-time optimizer allows us to simulate and optimize plant operations, ensuring maximum efficiency and optimal yield.
In addition, we have implemented AI-based predictive analytics for equipment maintenance. This helps us minimize unplanned downtimes by providing early alerts on potential failures. While scheduled maintenance is a given, AI ensures that critical assets are monitored continuously, helping us proactively mitigate risks.
Safety and reliability are also key priorities, especially given that we operate in a hazardous environment with complex processes. By integrating various data sources—including digital logbooks, e-permits, quality parameters from e-LIMS, and water information management systems—into an AI-driven platform, we generate real-time heatmaps and predictive insights. Our MQ AI platform helps identify potential safety risks, allowing us to take preventive measures before an incident occurs.
AI, therefore, plays a dual role: ensuring compliance and providing forward-looking insights for complex operations.
CIO&Leader: Integrating legacy systems remains a significant challenge in scaling AI effectively. What best practices should enterprises follow to overcome this issue?
Sumit Duttagupta: That’s a very real challenge. Without a well-structured and clean foundational database, AI will produce inaccurate insights. Enterprises must focus on data normalization and standardization before layering AI on top.
When integrating AI with multiple siloed systems, a clear strategy for data interfacing is crucial—particularly in defining primary and secondary keys and building real-time data caches. Additionally, contextualizing machine-generated data with business data is essential. If the integration lacks context, AI models will struggle to extract meaningful insights.
This requires a collaborative effort between IT, data scientists, and domain experts. For example, if I want AI to analyze pump performance in relation to our captive power plant, I need engineers who understand both the technical aspects and energy distribution patterns. A strong data governance framework, combined with domain expertise, ensures AI delivers relevant insights.
Another important factor is UI/UX. If AI-powered analytics tools aren’t intuitive, users won’t engage with them effectively. A simple, well-designed interface encourages adoption and maximizes the value derived from AI.
CIO&Leader: What are some of the AI-driven projects and innovations currently in progress at Haldia?
Sumit Duttagupta: Within our organization, most work processes are now digitized, providing us with a solid data foundation. We have adopted the ISA-95 framework, integrating automation at the root level, implementing Manufacturing Execution Systems (MES), and linking these to business insights.
Now, we’re focusing on leveraging AI agents. One key area of exploration is private LLMs tailored to our specific needs. Unlike generic LLMs like OpenAI’s models, which generate excessive and often irrelevant data, custom models would deliver more precise recommendations.
Other AI-driven initiatives include:
- Process Simulation for Polymer Production: AI helps us analyze cause-and-effect relationships when disruptions occur, enabling faster corrective actions.
- AI-Enhanced Demand Forecasting: By incorporating external factors such as shipping trends, geopolitical risks, and commodity price fluctuations, we aim to improve planning accuracy.
- AI-Generated Heatmaps for Plant Reliability: We have piloted an AI model for early event detection and are now scaling it across the plant.
- Customer Behavior Analytics: With over 4,500 customers, AI-driven segmentation is helping us optimize product distribution and pricing strategies.
- AI-Augmented Roles: We are exploring AI-assisted decision-making tools for procurement and HR performance management, aiming to streamline operations and improve efficiency.
Some of these initiatives are in pilot stages, while others will be scaled in 2025, subject to budgeting considerations.
CIO&Leader: Is ROI a major concern for CIOs when implementing AI, and is there increasing pressure from the board to define clear performance metrics for its success?
Sumit Duttagupta: Absolutely, ROI is critical. No enterprise will adopt AI just for the sake of it. However, senior management is increasingly open to experimenting with AI through small-scale pilots. They recognize that AI adoption is an iterative process—initial projects may not yield full returns immediately, but incremental successes build confidence and justify larger investments.
Rather than deploying AI across all processes, organizations should identify revenue-generating or efficiency-enhancing use cases first. Once proven, these initiatives can be scaled.
CIO&Leader: What has changed in AI priorities for CIOs this year as compared to the last year?
Sumit Duttagupta: The biggest shift in 2025 is the move from AI concepts and prototypes to real-world applications. Senior management is looking for tangible, operational AI successes.
CIOs are now focusing on two key aspects:
- Leveraging proven AI models—rather than experimenting with untested solutions, organizations are evaluating AI implementations in peer industries and adapting them.
- Securing early wins—quick success stories help build confidence among decision-makers, leading to broader AI adoption.
CIO&Leader: Given budget constraints and the need for cost optimization, how should CIOs balance innovation with financial sustainability?
Sumit Duttagupta: Budget constraints will always exist, but this doesn’t mean CIOs should take a step back from innovation. What we need to do is thoroughly understand the domain alongside business stakeholders and identify two or three high-impact use cases that affect operations, production, or procurement – for instance, implementing a faceless purchase system.
While these initiatives may sound good in theory, the real challenge lies in implementation: how do you model these scenarios in an AI system and create solutions that work effectively on the ground? If you can select and implement the right use cases, the ROI will naturally follow. Initially, when building these systems, the ROI might be conservative, but the benefits become clear during implementation.
There are multiple approaches to achieving ROI. In some cases, you might have excess staff, so you can automate processes and redeploy personnel to other areas. In other cases, where you have limited manpower, you can make the existing processes more intelligent and retrain the small team to perform better analytics using the new system. Through these approaches, we’re improving process efficiency and demonstrating tangible value in terms of both productivity and financial returns.
CIO&Leader: Talent acquisition and skill gaps remain major barriers to AI adoption. What strategies can CIOs implement to address these challenges?
Sumit Duttagupta: AI isn’t just about coding. Successful implementation requires a cross-functional team, including domain experts, business leaders, and tech enthusiasts. When undertaking a business or digital transformation program where AI is central to the journey, it’s crucial to build a comprehensive team. This team should include domain experts and business professionals with deep operational experience who understand their areas inside and out.
Within your organization, identify technology enthusiasts who have shown initiative in adopting new tools and systems. These individuals can be valuable team members after some reskilling in new technologies. Once you’ve assembled and prepared this internal team to handle scope understanding and deployment of AI solutions, partner with the right consulting firm that brings both technological expertise and global best practices.
With this alliance in place, begin working on specific use cases. Build these systematically, gain momentum, and establish confidence among both senior management and end users. Only then should you embark on a broader digital transformation journey that can effectively impact the entire organization.
To summarize, organizations should:
- Identify tech-savvy employees and upskill them in AI.
- Collaborate with consultants to bring in global best practices.
- Start with targeted AI use cases and gradually scale.
By developing an AI-focused culture, CIOs can ensure sustainable digital transformation.