AI on the Factory Floor: How India's COOs Are Deploying Intelligent Operations at Scale

In Tata Steel's Jamshedpur works — a manufacturing complex that has been producing steel for over a century — an artificial intelligence system now monitors thousands of sensor data points across blast furnaces, rolling mills and finishing lines in real time, predicting equipment failures before they occur and suggesting operational adjustments that reduce energy consumption and improve yield. The system, developed in partnership with a European industrial AI specialist and customised to Tata Steel's specific operational parameters, has reduced unplanned downtime by over 25% in the facilities where it has been fully deployed.

This is not a pilot. It is not a proof of concept. It is operational AI at the scale of one of the world's largest integrated steel plants — and it is just one example of a transformation that is accelerating across India's industrial base in 2025.

The question that India's COOs face is no longer whether to deploy AI in their operations. That question was settled in the boardrooms and strategy sessions of 2022 and 2023. The question now is how: how to scale AI from the isolated pilots that most large Indian manufacturers have been running for three to five years into enterprise-wide intelligent operations that deliver sustained competitive advantage.

The Landscape of Industrial AI in India

India's industrial AI adoption is at a critical inflection point. McKinsey's India Manufacturing Insights 2024 report estimates that Indian manufacturers have collectively spent over ₹12,000 crore on Industry 4.0 and AI-related technology investments in the past four years, with automotive, steel, cement and consumer goods leading the adoption curve. Yet the same report notes that only approximately 18% of these investments have moved beyond pilot stage to full-scale deployment — a gap between ambition and execution that represents both a challenge and an opportunity.

The reasons for this pilot-to-scale gap are instructive. They include data quality issues — legacy manufacturing equipment that generates incomplete or inconsistent sensor data — organisational resistance from operations staff who perceive AI systems as threats to their expertise and autonomy, IT-OT (information technology and operational technology) integration complexity, and a shortage of people who combine deep manufacturing domain knowledge with AI and data engineering capability.

For the COO, each of these is an organisational problem to be solved, not merely a technical one. And the COOs who are solving them are the ones whose organisations are now operating at a fundamentally different level of efficiency from their peers.

Predictive Maintenance: From Reactive to Anticipatory

Predictive maintenance — the use of sensor data, machine learning models and edge computing to predict equipment failures before they cause unplanned downtime — is arguably the most mature and most widely deployed AI application in Indian manufacturing. The business case is straightforward: unplanned downtime in a large Indian steel plant, petrochemical complex or automotive assembly facility costs between ₹1 crore and ₹10 crore per hour, depending on the asset. A predictive maintenance system that reduces unplanned downtime by even 20% generates a return on investment that is difficult to argue against.

Larsen & Toubro's heavy engineering division has deployed predictive maintenance across its turbine and compressor manufacturing operations, using vibration analysis and thermal imaging data to predict bearing failures and seal degradation with lead times of three to seven days. This window allows maintenance to be scheduled during planned production breaks rather than scrambled in response to failures — a shift that reduces both direct maintenance costs and the secondary costs of production disruption.

Mahindra's automotive manufacturing operations have implemented a predictive maintenance programme across their Pune and Nashik plants that monitors over 2,000 pieces of equipment. The system uses a combination of IoT sensors, historical maintenance records and AI models trained on failure data from comparable equipment globally to generate daily prioritised maintenance recommendations. In the first eighteen months of full deployment, unplanned downtime fell by 31% and maintenance labour costs declined by 18%, as reactive emergency repairs were replaced by planned preventive interventions.

The data challenge: Implementing predictive maintenance at scale requires solving a data problem that most Indian manufacturers have not yet fully addressed. Legacy equipment — machinery that may be ten, twenty or thirty years old — typically lacks the native sensor instrumentation that modern predictive maintenance systems require. Retrofitting sensors to legacy equipment is technically possible but operationally complex and expensive. Several Indian companies have addressed this through a combination of low-cost IoT sensor kits and edge computing platforms that can extract useful data from older machines — an approach that Pune-based industrial IoT specialist Altizon has pioneered and deployed across several large Indian manufacturing clients.

"The COO who waits for their entire asset base to be new and sensor-rich before deploying predictive maintenance will wait forever. The skill is retrofitting intelligence onto legacy operations while building the data foundations for the next generation." — Chief Operating Officer, large Indian industrial conglomerate, Gladwin International COO Forum, June 2025.

Machine Vision and Quality Intelligence

Quality control is the second major frontier of AI deployment in Indian manufacturing. Traditional quality inspection in Indian factories has relied on manual visual inspection — a process that is labour-intensive, inconsistent, and increasingly difficult to staff as the labour market tightens. Machine vision systems — cameras, lighting systems and AI models trained to detect specific defect types — offer a scalable alternative that is faster, more consistent and continuously improvable.

India's pharmaceutical sector has been particularly aggressive in machine vision deployment, driven by the exacting quality requirements of US FDA and European EMA regulatory compliance. Sun Pharma, Dr Reddy's and Cipla have deployed machine vision systems in their tablet and capsule inspection lines that can examine 200,000 to 400,000 units per hour with defect detection accuracy exceeding 99.9% — a standard that no manual inspection process can match at comparable speed.

In automotive, Bajaj Auto's Chakan facility has implemented AI-powered visual inspection at multiple assembly and finishing stages. The system uses convolutional neural networks trained on thousands of images of both acceptable and defective components to detect surface defects, assembly errors and dimensional deviations with millisecond-level response times. The technology has enabled a 40% reduction in quality-related warranty claims on exported vehicles — a direct contribution to Bajaj's export competitiveness in markets from Africa to Latin America.

The competitive implications of machine vision quality systems extend beyond the direct cost savings from reduced defects and warranty claims. They change the information environment for quality management. Traditional manual inspection generates aggregate pass/fail data. Machine vision generates granular defect location, type and frequency data that, when analysed over time, reveals patterns in the production process that cause defects — enabling root-cause elimination rather than just defect detection.

Autonomous Logistics and Warehouse Intelligence

The movement of materials within manufacturing facilities — the intra-logistics of raw material staging, work-in-process movement and finished goods handling — is a high-cost, labour-intensive activity that is increasingly being transformed by autonomous systems. Autonomous mobile robots, automated guided vehicles and AI-powered warehouse management systems are reshaping the intra-logistics function at India's most advanced manufacturing sites.

Godrej & Boyce's manufacturing complex in Vikhroli, Mumbai has deployed a fleet of autonomous mobile robots in its appliance manufacturing operations, handling the movement of components between production stages. The robots navigate dynamically, avoiding human workers and responding to real-time production scheduling changes — a capability that was impossible with the fixed-route automated guided vehicles of previous generations.

In the pharmaceutical logistics sector, Divi's Laboratories has implemented an AI-powered automated storage and retrieval system at its Hyderabad facility that manages over 40,000 stock-keeping units with near-perfect inventory accuracy. The system interfaces directly with the company's ERP and order management systems, enabling same-day fulfilment of complex multi-SKU orders that previously required a day and a half of manual picking and preparation.

The COO as Chief AI Officer for Operations

The deployment of AI in manufacturing operations at the scale and sophistication described above is fundamentally changing the COO role. In the most advanced organisations, the COO is effectively functioning as the chief AI officer for operations — responsible not just for the deployment of specific AI applications, but for building the data infrastructure, the technology architecture and the organisational capability that allows AI to be applied continuously across the operational landscape.

This demands a new set of competencies from operations leaders. The ability to evaluate AI vendor claims critically — to distinguish genuine operational AI capability from marketing-enhanced demos — is increasingly important as the industrial AI market becomes crowded with providers of varying quality. The ability to lead cross-functional AI implementation projects that bridge OT and IT teams, operations staff and data scientists, requires facilitation and translation skills that go well beyond traditional operations management.

Perhaps most importantly, the COO must manage the human dimension of AI deployment in manufacturing — addressing the legitimate concerns of operations staff about job displacement, building the trust that encourages front-line workers to engage with AI-generated insights rather than dismissing or circumventing them, and developing new roles that combine domain expertise with AI tool proficiency.

Hiring implications: Gladwin International is seeing a significant evolution in COO job specifications among India's leading manufacturers. The ability to lead AI-enabled operational transformation is now listed as a core requirement — not a nice-to-have — in over 70% of the COO mandates we received in the first half of 2025. Candidates who can demonstrate concrete AI deployment results — specific applications deployed, scale achieved, measurable operational outcomes delivered — command a significant premium in the talent market.

The executives most in demand combine genuine manufacturing operations experience with demonstrated capability in data-driven management. Pure technologists who lack operations credibility struggle in COO roles that require commanding the respect of experienced shop-floor leaders. Conversely, operations veterans who have not made the personal investment in understanding AI tools and their operational implications are finding themselves at a growing disadvantage in a market where the COO brief is being rewritten in real time.

The window for operations leaders to build this capability is finite. The organisations that have already deployed AI at scale are generating operational insights and building institutional capability that will compound over time. The gap between AI-enabled and AI-laggard manufacturers is widening. For India's COOs, the imperative to move from pilot to production is not merely a technology decision — it is a competitive one.