Resources and content
Use cases September 9, 2026 17 min read

OT/IT integration: why connecting systems is not the same as integrating data

An integration can work technically and still produce an architecture that is hard to govern, maintain and scale.

Athenas — OT/IT integration: why connecting systems is not the same as integrating data
In this article

There is a practical difference between two statements that are often treated as equivalent: “the systems are connected” and “the data is integrated.” The first describes connectivity — there is a path for information to travel through. The second describes an architectural capability — information reaches the consumer with meaning, context and traceability, and new consumers can use it without redoing the path.

An integration can work technically for years and, at the same time, make the architecture harder to govern. This article details why that happens in OT/IT environments and what changes when an organization starts treating industrial data as a reusable layer.

OT and IT do not necessarily speak the same language

Two environments, two sets of systems

OT — operational technology

  • Sensors and instrumentation
  • PLCs
  • SCADA
  • DCS
  • Process historians
  • Machines and equipment
  • Control systems

IT — information technology

  • ERP
  • CRM
  • Corporate databases
  • Analytics platforms
  • Various business systems
  • Data platforms and data lakes
  • Business applications

The gap between the two environments is not just about technology. It involves different operational priorities and different ways of treating risk, time and change.

Differences that affect integration
DimensionOT environmentIT environment
CriticalityDowntime can stop production or affect safetyDowntime affects administrative processes and analyses
AvailabilityContinuous operation, rare maintenance windowsMore frequent and negotiable maintenance windows
TimingMilliseconds to seconds, with continuous seriesTransactions, batches and periodic closings
ProtocolsOPC UA, Modbus, MQTT, proprietary protocolsREST, SQL, corporate messaging, files
SecurityNetwork segmentation, physical and logical access controlCorporate identity, IT policies, auditing
GovernanceOwned by engineering and automationOwned by IT and data teams
Life cycleEquipment and systems running for decadesShorter upgrade cycles
Data modelsTags, addresses and control-oriented structuresBusiness entities, tables and documents

Integrating OT and IT means dealing with these differences explicitly. When they are not addressed in the architecture, every project ends up solving the same set of problems its own way.

The problem with point-to-point integration

The most direct way to meet a need is to connect whoever produces the data directly to whoever needs it. Each connection, in isolation, makes sense.

Illustrative architecture

SCADA → MES

Supervisory data feeding production tracking.

SCADA → BI

Parallel extraction for operational reports.

MES → ERP

Production records reaching the corporate system.

MES → BI

Production indicators replicated for analysis.

ERP → Data Lake

Periodic loads into the analytics platform.

Historian → BI

Time series exported for reporting.

Historian → AI

Dedicated extraction for a modeling project.

Not every architecture needs to connect every system to every other, and there are cases where a direct integration is the right choice. What the example shows is the pattern of growth: every new source or consumer tends to add new connections, and every connection adds its own transformation, scheduling, error handling and maintenance.

When an integration works, but the architecture gets worse

  • Dependencies: a consumer becomes dependent on an industrial source's internal format.
  • Duplicated code: the same read is reimplemented across different projects.
  • Different transformations: two reports calculate the same indicator in diverging ways.
  • Scattered rules: limits, conversions and filters end up spread across scripts and jobs.
  • Hard to trace: it becomes impossible to say with confidence where a number came from.
  • Concentrated knowledge: only whoever built the integration knows how it behaves.
  • Growing maintenance: every source change requires review in multiple places.
  • Hard to change: renaming an asset or changing the hierarchy becomes a project.
The side effect of the local fix

Local problem solved

The application that needed the data started receiving it. The request was met on time, with a small scope and without involving other areas.

Architectural complexity created

The organization now has one more dependency between source and consumer, one more transformation rule outside the common model, and one more point that needs to be reviewed with every change on the plant floor.

Connecting is not contextualizing

Consider the same piece of information as seen by three different systems in a hypothetical operation.

The same reality, three representations

SCADA

  • Tag = LT_302
  • Value = 71.3

ERP

  • Equipment = Tank 302

MES

  • Unit = Plant A
  • Line = Process 03

Moving the value 71.3 into an analytics database solves transport, not understanding. Questions that any consumer needs to ask remain unanswered:

  • What is LT_302?
  • Which equipment does this tag represent?
  • Where is that equipment located?
  • Which process uses it?
  • What is the unit of measure for the value?
  • Which product was being processed?
  • Is this data valid?
Integration, however, is only part of the problem. Data also needs to carry operational context.

The architecture needs to know the structure of the operation

Operation hierarchy
  1. 01GroupA set of plants under the same organization.
  2. 02PlantAn industrial unit, with its own operational reality.
  3. 03AreaA functional section of the plant, such as reaction, compression or utilities.
  4. 04LineA production sequence within the area.
  5. 05EquipmentA specific asset, with a stable identifier.
  6. 06VariableA measured or calculated quantity of the equipment.
What each level needs to carry

Variable

  • Value
  • Timestamp
  • Unit of measure
  • Quality
  • Origin

Equipment

  • Process
  • Product
  • Order
  • Maintenance
  • Events

When this structure exists in the architecture, and not only in documents and in the teams' knowledge, data becomes locatable by what it represents. That is the change that lets a new consumer find the information without needing a source specialist.

Data-oriented architecture

Two ways to organize data projects

Traditional model

Each project creates its own integrations, with its own transformations, its own identifiers and its own interpretation of the data. The result gets delivered, but stays limited to the project that built it.

Data-oriented model

The organization builds a reusable data layer, with a declared information model, context and quality. Projects start consuming that layer instead of recreating it.

Structure of the data-oriented model
  1. 01SourcesIndustrial and corporate systems that produce the data.
  2. 02Data layerCollection, contextualization, quality, history and distribution under a common model.
  3. 03Multiple consumersBI, MES, analytics, applications, AI, maintenance, energy and quality all consuming the same base.

Where the Unified Namespace fits in

The Unified Namespace is a single, hierarchical structure where the state of the operation is published and consumed. Instead of each application knowing internal source addresses, it accesses a path that describes the operation.

Flow with a common layer
  1. 01SourcesPublish data with declared context.
  2. 02UNSHolds the structure, meaning and current state of the operation.
  3. 03BI / Analytics / AI / ApplicationsConsume the same structure, under defined access rules.

The goal is not simply to replace one integration model with another. It is to create a common structure in which consumers know where to find the data, what it represents, what its context is, and how it relates to other data in the operation.

What changes for engineering

The question engineering starts asking

Before

“I need to fetch this data from System X.”

After

“I need to consume this entity's contextualized data.”

Practical effects

Reuse

The same contextualized data serves different projects without redoing the interpretation work.

Governance

Responsibilities and permissions are defined over the model, not over scripts.

Documentation

The information model works as living documentation of what exists in the operation.

Observability

Collection failures and delays become visible in the chain before they affect decisions.

Standardization

Units, naming and quality rules are declared in a single place.

Decoupling

Source changes stop propagating direct breakage to every consumer.

Scalability

Adding consumers no longer means adding integrations at the source.

Practical example: a new application needs access to production data

Teaching example

New production-tracking application

A hypothetical company has multiple plants, supervisory systems, MES, ERP, historians and a BI platform. A need arises for a new application that tracks production by line, with product and shift context.

There are two possible paths, and both have consequences.

Fragmented approach
Build new integrations that fetch information directly from each source. Fast delivery on the first scope, with more dependencies, more duplicated rules and more maintenance points.
Common-layer approach
Consume data already organized and contextualized in the data layer. Requires the modeling to exist, and pays back with reuse for the next applications.

The second approach is not free: it shifts effort to modeling, governance, security, quality, history and ownership definition. The difference is that this effort is done once and serves the following projects, instead of repeating with every new request. Hypothetical scenario, for illustration purposes.

The role of Athenas

Athenas Data Intelligence works by building a layer of contextualized industrial data, connecting heterogeneous sources to a common structure that different applications can consume.

Athenas' position in the architecture
  1. 01OTSensors, PLCs, SCADA, historians and existing industrial systems.
  2. 02AthenasCollection, catalog, normalization, ISA-95 UNS, historian, distribution routes, observability and governance.
  3. 03Contextualized dataValues with declared asset, hierarchy, unit, quality and origin.
  4. 04BI / Analytics / AI / ApplicationsConsumers reading from the same base, through a semantic path.

Two caveats matter: Athenas does not replace industrial systems — SCADA, MES and ERP keep performing their roles — and it does not eliminate every integration. What changes is where they happen and how many times they need to be redone.

Checklist for a sustainable OT/IT architecture

OT/IT checklist
  • Industrial sources mapped
  • Corporate systems identified
  • Hierarchy defined
  • Information model defined
  • Standardized naming
  • Context attached to the data
  • Quality monitored
  • Traceable origin
  • Security defined
  • Consumers decoupled from sources
  • Observable architecture
  • Governance established

Integration is an architectural capability, not just a connection

Connecting two systems can solve a specific need. Building an integrated data architecture requires a different concern: creating a structure where new applications can consume information without continuously reproducing the complexity of the original sources.

It is this shift, from isolated integrations to a contextualized, reusable data base, that allows the use of industrial data to scale.

EAEquipe AthenasAthenas Data Intelligence
Share

See how Athenas organizes your operation's data

A roughly 40-minute demo showing source connectivity, the catalog, the ISA-95 UNS and real-time data distribution.