The Intelligence System for Industrial Assets

Why manufacturers need a new intelligence layer for their industrial assets.

INTRODUCTION

Executive summary

Manufacturers have invested heavily in automation, data infrastructure and specialised software, yet engineers still have to bring information together manually to understand what is happening within their assets. Data sits across historians, production systems, maintenance applications and technical documentation, while critical operational knowledge remains with individual experts. The result is an industrial intelligence gap between the information manufacturers possess and their ability to turn it into understanding, decisions and action.

Marentis is building The Intelligence System for Industrial Assets to close this gap. It connects operational data, technical knowledge, specialised analytical methods and accumulated experience around a persistent understanding of the asset. Instead of leaving engineers to combine dashboards, models, documents and previous investigations themselves, Marentis brings evidence together in one workflow, keeps conclusions traceable and turns the outcome of each investigation into reusable operational knowledge.

In an active deployment with a large automotive manufacturer, investigations that previously took several days were reduced to a matter of hours, while participating production and maintenance engineers reported an overall 30 percent productivity gain in their daily work.

Marentis' mission is bringing the AI revolution to the machines that drive the physical world, empowering every engineer and operator to achieve more. Its vision is an intelligence layer through which assets can explain their condition, preserve what the organisation learns about them and progressively support the work required to operate and improve them.

Over time, this foundation enables a controlled progression from assisted intelligence to agentic and autonomous intelligence. The long-term goal is metacognitive industrial intelligence: a system that understands what it knows about an asset, recognises uncertainty, learns from previous decisions and outcomes, and progressively takes responsibility for defined operational tasks within explicit human, safety and governance controls.

For manufacturers evaluating Industrial AI, the relevant test is not the sophistication of an individual model or demonstration. The system should improve real operational work, fit existing industrial environments, keep conclusions traceable to evidence and build reusable asset knowledge that becomes more valuable over time.

SECTION 1

The industrial intelligence gap

Manufacturing has become highly automated, but the way people understand and work with industrial equipment remains largely unchanged.

A modern plant may contain thousands of sensors, PLCs, robots, controllers, cameras and software systems. Operational data sits in historians, SCADA systems, MES platforms, quality databases and maintenance applications. Manuals, drawings and service instructions are stored elsewhere. Engineers and operators add another source of knowledge through their experience with the equipment.

When an asset behaves unexpectedly, someone still has to connect these sources. An engineer may begin with an alarm, inspect telemetry, review maintenance history, search the machine manufacturer's manual and ask whether the issue has occurred before. An engineer may also look at analytics tools built internally or by vendors, spending additional time interpreting the output, identifying a root cause and turning it into a concrete action. It is this last step that matters most. The answer to an issue may be found, but the process is slow, difficult to repeat and dependent on a small number of experienced people.

This is the industrial intelligence gap: the difference between the information a manufacturer has and its ability to use that information to understand what is happening, determine why it is happening and decide what to do next.

More data has not created a shared understanding

The underlying problem is fragmentation. Production, quality, maintenance and engineering are commonly supported by separate systems with different data structures and responsibilities. NIST continues to identify industrial data complexity, system integration and usable context as barriers to the broader use of AI in manufacturing.

The same problem exists within an individual asset. A temperature value has limited meaning without knowing the operating state, product variant, process step and recent changes to the machine. An alarm becomes more useful when it is connected to the relevant component, surrounding signals, technical documentation and previous incidents.

Most manufacturers already hold much of the information needed to understand an issue. What they lack is a way to bring the relevant information together when it is needed.

Expert capacity is spent rebuilding context

Across Marentis customer conversations and deployments, manufacturers estimate that experienced engineers spend between 20 and 30 percent of their working time searching for data, reconstructing events and gathering the context needed to investigate operational issues. That is roughly one to one and a half working days per engineer each week.

This time goes into finding the right signals, comparing information across systems, locating relevant documentation and previous incidents, and assembling the context needed before analysis can begin.

When the required knowledge is not available internally, manufacturers depend on machine suppliers, system integrators and external service specialists. In Marentis customer conversations, external industrial specialists are commonly quoted at between 1,500 and 2,500 euro per day, before travel expenses or minimum service commitments. External support will remain necessary, but manufacturers should not need it for questions that could be answered from their own data and knowledge.