Intelligent data pipelines for industrial operations: Introducing flows in CONNECT

Posted: June 10, 2026

At AVEVA World in Milan, we introduced flows as a new capability within CONNECT. What struck me in the conversations that followed was not the questions about features or pricing. It was how quickly people moved to describing their own situation. Data stuck between OT and IT systems with no clean path between them. Streams arriving unstructured, without the context to make them useful. Custom integrations built over years that users are now struggling to maintain. The same problems, from different industries, different geographies and across different stages of the CONNECT journey.

Those problems have a common root. Industrial data is being generated continuously, but between the machine and the decision, it loses its timing, its structure and its meaning. What is missing is intelligence within the data pipeline itself. Flows were conceptualized to close exactly that gap.

Flows are CONNECT's real-time data pipelines, built for data in motion. The name is deliberate. The Crosser team has spent nearly eight years deploying flows and solving real integration problems across some of the most demanding industrial environments in the world. They called it flows because that is what industrial data does.

What are flows?

When a temperature spikes, a pressure drops or a line stops unexpectedly, that moment generates data. But machines generate raw numbers: a reading, a value, a signal. That number has no name, no unit, no relationship to anything around it. On its own it is meaningless. Stripped of context, it cannot tell you whether a threshold has been crossed, whether the reading is normal for that asset or whether anything downstream needs to respond.

Before data travels anywhere, flows do several things to it right at the source.

At the edge, flows filter out the noise so only meaningful streams move, put the data into a common format so different systems can read it, and add the context that makes a number useful: which machine, which line, which process. Flows also apply rules, so if something crosses a threshold, the right action triggers immediately, locally, at the edge.

What arrives is not raw data waiting to be processed. It is data that is already useful.

How you can deploy flows

Understanding how flows achieve this starts with how they are integrated within CONNECT. Flows operate through two components:

1. Flow manager sits inside CONNECT. It is where you design, deploy and monitor your pipelines from a single place, across every site and system.
2. Flow runtime sits wherever your data lives: at the edge, on premise or in the cloud. It is the engine that does the processing work at the source, before data travels anywhere.

Together they cover the full industrial and enterprise stack. On the operational side, that means sensors, PLCs, SCADA, DCS, MES and historians, including AVEVA™ PI System™. On the enterprise side, it means ERP, CRM, databases, cloud platforms and SaaS applications. With connectivity across 800+ systems, flows support multi-source, multi-destination pipelines—collected from any source, processed in motion, and delivered to any destination, in any direction.

For existing AVEVA customers, flows extend what CONNECT already does. Data can be collected from AVEVA™ System Platform, AVEVA™ PI Server or AVEVA™ Unified Engineering just as easily as it can be delivered to them. Once processed at the source, data moves directly to wherever it needs to go—into CONNECT, back to any AVEVA system or to any third party system in the tech stack. The destination is always your choice. What flows enables is the quality, context and timing of data streams, regardless of the destination.

Flows ingest the full industrial stack. On the factory floor, teams can connect PLCs, sensors, MES and historians into a single coherent data stream without building a custom integration for each system. At the edge, OT data from sensors, PLCs and operational systems can power real-time quality monitoring, anomaly detection and safety workflows, with logical execution locally before data travels anywhere. Across the enterprise, flows connect operational data to ERP, CRM and cloud platforms in any direction, supporting intelligent automations that span the full data value chain from factory floor to the boardroom. And for organizations moving toward a Unified Namespace architecture, flows provide the pipeline layer that publishes standardized data models continuously across sites and systems.

Build for everyone, not just data engineers

Flows are not built for data engineers alone. Flow studio's low-code environment means that anyone who understands the process, whether a controls engineer, an operations manager or a plant technician, can design, test and deploy a pipeline without any code. Flow studio provides over 200 ready-made modules covering transformation, enrichment, logic, analytics and integration, so teams are not starting from scratch. Pipelines can be built, tested and deployed in hours rather than months. Once live, users have access to full visibility across every flow, every runtime and every site from a single place inside CONNECT: events, logs and performance in one view. The same pipeline deployed onsite can be running across any number of sites without rebuilding it each time.

How flows support industrial AI

For organizations investing in industrial AI, the pipeline layer is where most projects quietly fail. Models are ready. Dashboards are built. But the data arriving at them is inconsistent, delayed or missing the context that makes a prediction meaningful. Flows address that problem in two ways. First, by conditioning data at the source so what arrives at CONNECT is already clean, timed and contextualized. Second, by supporting the execution of your own ML models inside the pipeline itself. Anomaly detection, predictive logic, custom calculations: teams can deploy their own models to run at the edge, inside the flow, before data travels anywhere. By the time it reaches CONNECT's analytics tools, the heaviest processing is already done. Flows conditions data and executes logic at the source, so downstream AI and analytics start from data that is already contextualized, normalized and ready to use.

This is what intelligent data pipelines mean in practice. Not an additional abstract feature or a connector, but intelligence applied to data while it is still moving, before it reaches the model, before it reaches the dashboard, before any decision is made. CONNECT brings together industrial data, analytics and applications across the full OT, ET and IT stack. Flows strengthen that foundation by ensuring that data streams arrive already prepared and already contextualized. For industrial organizations closing the gap between siloed and unstructured data generation and decision-making, that journey starts with intelligent pipelines. It starts with flows.