Halconia.
Let’s talk
Back to blog

Why integrate operational knowledge into industrial automation systems

Integrating operational knowledge into industrial automation allows operational experience to be not only recorded but also made accessible and reusable within…

Andrés Naves ·

Integrating operational knowledge into industrial automation allows operational experience to be not only recorded but also made accessible and reusable within the production system (KNOWLEDGE CAPTURE AS A CONTROL PROBLEM IN INDUSTRIAL OPERATIONS: A…). To decide the scope of your project, it is worth distinguishing between externalizing that knowledge, making it accessible, and enabling its reuse.

Start with the foundation in our previous article: How to implement governance in industrial AI systems | Halconia

Why integrate operational knowledge into industrial automation?

The reason for integrating operational knowledge is to preserve and reuse the learning from operations as part of the production system (LUMEN ET VIRTUS). In AI risk management, the context provided by experts and other relevant participants is used to reduce the likelihood of system failures and negative impacts (Artificial Intelligence Risk Management Framework (AI RMF 1.0)).

Purely manual capture tends to generate informational instability, low traceability and recurring loss of operational learning, according to the conceptual model published in LUMEN ET VIRTUS. That model approaches capture as a dynamic systems and control problem, rather than limiting it to digitization or knowledge management. Its approach places operational experience within a system whose behavior must be understood over time and under real operating conditions.

Integrated architectures favor the consolidation and reuse of knowledge as part of the behavior of the production system, within that theoretical-conceptual approach (LUMEN ET VIRTUS). The cited study adopts a theoretical-conceptual methodology (LUMEN ET VIRTUS); in this article we use it to explain the integration of operational knowledge, without promising quantified productivity improvements or the absence of failures.

The NIST framework contemplates that the contexts and needs of AI actors evolve over time (National Institute of Standards and Technology). To assess your project, we propose including the updating of the operational context among the criteria for deciding which knowledge to integrate.

At Halconia, we document and structure the reality of the organization before designing any system, without starting from preconceived solutions. We turn accumulated experience into documented, accessible and controlled systems. This approach includes detecting bottlenecks, informal dependencies and repetitive work that does not scale.

What does it mean to turn operational knowledge into part of the system, rather than just storing it?

In this article, we use a working definition: integrating operational knowledge means representing it so that it can be externalized, be accessible, and be reused in operations. Storing information addresses its recording, but the LUMEN ET VIRTUS conceptual model also contemplates its accessibility and reuse as states of the knowledge system (LUMEN ET VIRTUS).

The model allows these states and the disturbances affecting knowledge under real operating conditions to be analyzed jointly (LUMEN ET VIRTUS). Variability and unexpected events appear as operational disturbances, so the approach does not treat knowledge as a static representation of activity.

The conceptual model lists sensors, multimodal records, digital platforms, and AI assistance mechanisms as technical actuators (LUMEN ET VIRTUS). To delimit your project, we propose examining these means in relation to the externalization, accessibility, and reuse of knowledge under the operational disturbances described in the model.

The approach brings together foundations from cyber-physical systems, architectures with human interaction, adaptive automation, and cognitive support structures (LUMEN ET VIRTUS). The study concludes, within its theoretical-conceptual approach, that formalizing knowledge capture as a technical problem expands the capacity to analyze and design industrial self-learning solutions and strengthens the integration between human cognition, automation, and operational performance (LUMEN ET VIRTUS).

In our offering, Knowledge Systems refers to structured bases that capture tacit knowledge and turn it into organizational infrastructure.

Three states to distinguish knowledge capture from operational integration

Knowledge state

What it represents in the conceptual model

Externalization

The representation of knowledge constitutes a state of the knowledge system.

Accessibility

The availability of knowledge constitutes a state distinct from its externalization.

Reuse

The model contemplates reusing knowledge as part of the behavior of the production system.

Why don't data and AI alone replace the judgment of those who know the operation?

Data and AI do not by themselves replace operational judgment, because representing complex human practices through mathematical models can eliminate necessary context (National Institute of Standards and Technology). In this article, the considerations about model limits and human interaction apply to AI systems, not indiscriminately to all industrial automation.

Many data-based approaches try to convert observation and decision practices, both individual and social, into measurable quantities, according to NIST (National Institute of Standards and Technology). This representation may lose context and make it harder to understand impacts relevant to AI risk management. Therefore, having a quantifiable representation of a practice does not necessarily amount to preserving the context needed to understand it.

The NIST framework contemplates documenting the limits of the system's knowledge and how its outputs can be used and overseen by people (National Institute of Standards and Technology). That documentation provides sufficient information to help relevant participants make decisions and take subsequent actions. Before using an output from the system, we recommend consulting the documentation on its knowledge limits and on how it can be used and overseen.

Human participation also does not automatically make a decision correct: under certain conditions, interaction with AI can amplify human biases (National Institute of Standards and Technology). NIST places perception-based judgment tasks among those conditions. When these variations are properly considered when organizing human-AI teams, they can produce complementarity and improve joint performance.

The relationship between people and systems admits different configurations: autonomous decisions, decisions delegated to an expert, or use of AI as an additional opinion (National Institute of Standards and Technology). Some systems may not require human oversight, while others may specifically need it.

In Halconia, AI Agents for industrial processes designates agents trained specifically with real sector knowledge, internal criteria, and operational context.

Sources