Beyond the AI hype: Why heavy industry needs intelligence, not just data

Falkor is positioning digital twins as shared operational environments that help industrial teams analyse plant conditions and coordinate decisions.

India's industrial AI opportunity is moving beyond the question of how much data companies can collect to a more consequential one: how effectively can they use that information to make decisions across large industrial facilities?

For energy producers, refiners and other industries operating large and complex assets, the stakes are considerable. An equipment failure can disrupt production, an inefficient shutdown can inflate costs, and an incorrect recommendation in a safety-critical environment can have consequences extending well beyond the balance sheet. While companies have invested in digitising operations, much of the information required to make timely decisions remains distributed across engineering systems, maintenance records, process control networks and the experience of individual employees.

The next phase of industrial transformation, therefore, may depend less on deploying another AI tool and more on connecting these fragmented sources of information into a reliable operational framework.

Mint was in conversation with Pedro Alcântara, Executive Vice President and Chief Revenue Officer, Falkor, and Deepak Kumar Swain, Managing Director, Falkor India, to explore how Industrial AI and digital twins are transforming decision-making across India's energy and heavy industries.

That is the proposition behind Falkor, the industrial intelligence company formerly known as Kongsberg Digital. The company is sharpening its focus on AI-powered digital twins and integrated industrial workflows as it looks to expand its presence in India. Its April 2026 partnership with Wipro to bring AI-powered digital twin solutions to the energy and utilities sector is one example of the industry's growing emphasis on combining engineering expertise, operational data and AI in industrial environments.

For Falkor, the distinction between conventional enterprise AI and industrial intelligence lies in the context in which the technology operates and the consequences of getting a decision wrong.

"It's moving from data to context. That's the journey when it comes to what industrial intelligence becomes," says Alcântara.

Companies, he argues, have made substantial progress in collecting and organising information. The challenge now is that operational knowledge remains fragmented across systems and, in many cases, inside the heads of experienced employees.

"It's very difficult to just transfer that knowledge if that knowledge is not recorded and if that knowledge is not brought forward in context with all the information that takes place from different operations in the organisation," he says.

The problem is particularly pronounced in industries operating complex assets, where an engineer investigating equipment performance may need to consult process data, maintenance histories, engineering drawings and inspection reports before arriving at a decision.

Connecting the dots across industrial operations

Falkor's approach centres on digital twins, virtual representations of physical assets that bring together live operational data, engineering models and enterprise information. Rather than treating a digital twin as a visualisation tool or a standalone dashboard, the company positions it as a shared operational environment through which teams can understand plant conditions, assess risks and coordinate work.

Swain says the challenge in heavy industries is to make AI work within the constraints of real-world operations.

"In asset-heavy industries, plant operations are governed by rigid safety limits where unexpected failure leads to serious financial and safety fallout," he says.

Falkor's Industrial Intelligence platform connects information that typically sits in different systems: process telemetry in distributed control systems (DCS) and supervisory control and data acquisition (SCADA) historians, maintenance records in enterprise resource planning systems, and asset designs in engineering drawings and three-dimensional models.

The resulting digital twin provides a contextualised view of plant operations. Machine learning can then be combined with engineering principles and physics-based simulations to help teams identify potential problems and evaluate possible interventions.

The distinction matters because industrial AI cannot be judged by the same standards as a consumer chatbot. In a consumer application, an inaccurate response may be inconvenient. In a refinery or a processing plant, a flawed recommendation could affect equipment integrity, production continuity, or worker safety.

"In our industry, if you are 99% right, you are 100% wrong," Alcântara says, underlining the importance of reliability in environments where decisions can have serious consequences.

For Swain, the answer is to ground AI in validated operational data and engineering context, allowing plant personnel to understand how a recommendation was reached rather than relying on an opaque output. Explainability and auditability become central to adoption, particularly where operators must retain control over safety-critical decisions.

Swain says this combination of operational context and engineering-led intelligence can help plant teams anticipate equipment issues, coordinate turnaround work and stabilise production.

From faster analysis to fewer disruptions

The commercial case for industrial AI is ultimately measured in operational outcomes: less downtime, more efficient maintenance, improved throughput and better use of engineering resources.

Alcântara cites examples from Falkor's customer engagements in which connecting systems and inspection information has substantially reduced the time required to prepare assessments. In one case involving the review of 1,000 inspection points, a process that previously took four to five days could be completed in three to four hours, he says.

The Vår Energi use case, which explores industrial digital twin readiness in greater detail, offers further context on how connecting operational information can support more effective industrial decision-making.

Alcântara also describes customer operations where centralising engineering systems, operating data and industrial knowledge has accelerated decisions associated with plant shutdowns and restarts. According to him, the speed of such decisions improved by about 70% in the examples he cited.

These are company-reported examples rather than industry-wide benchmarks. Their significance, however, extends beyond the time saved. Bringing multiple sources of information into a common context can help engineers consider more variables before making a decision.

"The speed, of course, helps a lot. Accuracy is the accuracy of the decision that took into consideration a situation awareness far beyond what that operator had access to before," Alcântara says.

In India, the opportunity spans several operational priorities. Swain identifies predictive asset reliability as one of the most immediate applications, with live condition monitoring potentially helping companies move away from rigid, calendar-based maintenance schedules towards interventions based on equipment health.

A second priority is turnaround optimisation. Planned shutdowns and restarts can involve multiple contractors, work permits, equipment dependencies, and tightly sequenced activities. Interactive digital twins can give teams a shared view of the work and help coordinate execution.

A third is the preservation of institutional knowledge. As experienced engineers retire or move on, their understanding of equipment behaviour, past failures and maintenance practices can be difficult to transfer to younger employees. Industrial AI tools can help make historical maintenance logs, equipment drawings and vendor manuals more accessible to technicians working in the field.

The underlying objective is to reduce the time employees spend searching for information and increase the time they spend resolving problems.

"People are becoming more valuable because they're making better decisions and they're spending less time doing work that systems can do for them," Alcântara says.

Why India matters

India's expanding energy and industrial base gives the transition particular significance. For operators managing complex assets, AI adoption is not simply about experimenting with new technology. It is also about improving productivity, extending asset life, managing energy consumption and strengthening operational resilience.

Swain says Indian energy producers and refiners are beginning to move away from isolated proofs of concept towards more integrated operational applications. The opportunity, he argues, is to connect data, engineering models and AI directly to everyday plant workflows rather than adding another layer of technology that employees must learn to navigate.

Alcântara sees India's combination of industrial expansion and digital capabilities as a strategic advantage.

"There are very few markets in the world that combine energy, manufacturing and the digital capacity that India has," he says.

He believes India can also approach industrial intelligence differently from markets where digital systems are being retrofitted into facilities originally designed for largely manual operations.

"I think India has the opportunity to make digital from the start and intelligent from the start," Alcântara says.

The company's own India operations reflect this evolution. According to Alcântara, the Bengaluru office initially focused primarily on product management, development and supporting international customers. It has since expanded its commercial focus to engage with Indian customers and participate in local tenders.

"That's the biggest change," he says. "The market became so relevant that the same organisation now is building product and engaging with customers locally."

Swain says the India team is also contributing to Falkor's global product strategy. Its Bengaluru-based engineers work on platform infrastructure, industrial contextualisation, deterministic simulation modelling and agentic AI workflows for deployment across international asset bases.

This positions the Indian operation as more than a development centre. It is part of the company's effort to build and scale the underlying capabilities required for industrial intelligence.

Swain describes a hybrid deployment approach in which real-time process control, raw sensor streams and safety-critical information can remain within plant networks or local private data centres, while aggregated analytical workloads can operate on sovereign cloud infrastructure located in India. Network segmentation and controlled data flows are intended to limit exposure of operational technology to external networks.

Such architecture is important because industrial systems are not interchangeable with ordinary enterprise applications. Connecting operational technology with information technology can create new opportunities for analysis, but it also requires careful management of access, system boundaries and the risks associated with unauthorised commands.

The human dimension is equally important. AI recommendations may help operators assess conditions and identify possible interventions, but the level of autonomy granted to a system must reflect the consequences of its decisions.

Alcântara argues that companies should distinguish between decisions that can safely be delegated to machines and those that require human judgement.

For him, trust must develop through validated use cases and a clear understanding of how a system reaches its conclusions. Rushing to remove human oversight, he cautions, could undermine adoption rather than accelerate it.

The challenge is organisational, not just technological

Despite the promise of industrial AI, deploying software alone will not transform plant operations. Companies must first establish which business problems they want to solve, how their teams should work together and what information is required to support those decisions.

Alcântara warns that technology can amplify existing inefficiencies if organisations fail to address fragmented workflows before introducing AI.

"If you don't organise your operating conditions before you bring this intelligence layer on top, then what AI will do is it will scale confusion," he says.

That is also why the shift from pilot projects to sustained deployment is likely to depend on organisational readiness as much as technical capability. A successful system must fit into existing workflows, make information accessible without creating additional administrative burdens, and give employees enough visibility to trust its recommendations.

For Swain, the next three to five years will see industrial intelligence increasingly applied to asset reliability, turnaround execution, and the transfer of technical knowledge across generations of engineers.

For Alcântara, the broader objective is to create a shared operational environment in which companies can learn from their own data and experience faster than their businesses grow.

As India expands its industrial capacity, the competitive advantage may not come simply from having more data or deploying more AI. It may come from connecting engineering knowledge, operational information and human judgement in ways that allow industrial organisations to make better decisions, safely, consistently and at scale.