The factory that understands itself
A modern manufacturing plant never stops generating information. Machines report their status, sensors monitor movement, cameras inspect products, mobile devices record frontline activity, and software captures every transaction across production, warehousing and logistics. Yet despite this constant flow of data, many manufacturers still struggle to answer one fundamental operational question: what is happening on the factory floor right now, and what should happen next?
For Stephan Pottel, Manufacturing Strategy Director EMEA at Zebra Technologies, that challenge reflects the limitations of thinking about digital transformation purely in terms of connectivity. Connecting machines and collecting more data have become essential foundations, but they are no longer enough on their own. Instead, manufacturers are beginning to combine real-time data capture, AI and automation to create what he describes as “ambient intelligence” – operational environments that continuously understand changing conditions, provide context for frontline workers and increasingly trigger appropriate actions without waiting for manual intervention.
The distinction matters because the role of manufacturing data has changed. For much of the Industry 4.0 era, organizations focused on making assets visible and connecting operational technology with enterprise systems. Those investments created unprecedented levels of transparency, yet visibility alone rarely improves productivity, quality or responsiveness. Manufacturers still need to interpret information, coordinate decisions across multiple functions and respond quickly enough to influence operational performance.
“Ambient intelligence brings together the physical layer, the data layer, an AI analysis layer and the execution layer,” Pottel explains. “It’s a living system or reflection of what’s happening in that moment. It connects the frontline, understands the whole environment, makes every asset visible and supports intelligent automation.”
That evolution is changing the role of factory architecture. Rather than simply moving data between machines, enterprise systems and the cloud, manufacturers are creating environments where information is captured continuously, interpreted in real time and delivered to the people, systems or AI agents best placed to act upon it. The objective is no longer simply to digitize operations. It is to build factories that are increasingly aware of their own operating conditions, enabling faster decisions, greater resilience and ultimately more autonomous manufacturing.
Context matters more than connectivity
The challenge facing many manufacturers is not a shortage of operational data but the difficulty of turning fragmented information into a coherent picture of what is happening across the business. Production systems, warehouse management platforms, manufacturing execution systems, enterprise applications and frontline devices all generate valuable data, yet too often that information remains fragmented across separate systems. Even where organizations have invested heavily in digital infrastructure, operational intelligence can remain incomplete, delayed or inconsistent.
The issue, he argues, is less about collecting additional information than creating an architecture capable of connecting it. “Manufacturers generate data across operational technology, edge workflows, manufacturing execution systems, warehouse management systems and central ERPs,” he says. “The challenge is how to connect the data between these operations with an architecture that’s secure and scalable.”
That complexity has accumulated over many years. Manufacturing sites frequently develop their own applications, data structures and integration methods to meet local operational requirements. While those solutions may perform well individually, they often create barriers when organizations attempt to establish a consistent operational view across multiple plants, business functions or supply chain partners. Data ownership becomes fragmented, formats differ and duplicate or obsolete information undermines confidence in the insights being generated.
Perhaps the biggest challenge is that many manufacturers continue to rely on systems designed to record events rather than support decisions as events unfold. Data may be updated every hour, overnight or at the end of a production shift, providing an accurate historical record but offering limited value when operational conditions are changing minute by minute.
He describes this as the difference between “systems of record” and “systems of reality.” Manufacturers increasingly need architectures capable of supporting event-driven operations, where information is captured, processed and shared as activities occur rather than according to scheduled reporting cycles. “Manufacturers want real-time intelligence rather than hourly, daily or weekly scheduled data updates,” he explains. “That means event-driven architectures and APIs become essential.”
Zebra’s research with Oxford Economics illustrates how difficult that transition remains. More than half of manufacturers identify legacy technologies as the single biggest barrier to improving workflows, ahead of concerns such as cybersecurity, training costs and executive alignment. Most also acknowledge that structured data analysis remains confined to individual functions or isolated projects rather than supporting decision-making across the business. The manufacturers making the greatest operational progress are those that have invested in integrated, high-quality data management capable of delivering trusted intelligence wherever decisions need to be made.
Intelligence belongs where the work happens
The discussion around manufacturing AI often centers on cloud platforms and increasingly powerful foundation models, yet many of the decisions that determine operational performance cannot wait for information to travel to a remote data center and back. Detecting a quality defect, preventing a safety incident or identifying an equipment anomaly frequently demands an immediate response. In those situations, milliseconds matter.
That is changing the way manufacturers think about where intelligence should reside. Rather than viewing the edge simply as a data collection point, organizations are beginning to deploy AI directly alongside the people, machines and workflows generating the information. The result is faster decisions, lower operating costs and greater operational resilience, particularly where production continuity depends on real-time action.
Advances in hardware, he argues, are accelerating that shift. “Modern mobile computers and machine vision hardware are engineered with neural processing units and CPUs for on-device AI inference,” he says. “We’re seeing this important strategic shift towards on-device AI models. Smaller models on mobile computers, tablets and vision controllers don’t need to send tokens to the cloud, so there’s a compelling financial case as well as an operational one.”
Running AI at the edge addresses two of manufacturers’ biggest concerns simultaneously. It reduces latency while allowing sensitive operational data to remain within the factory environment. It also avoids the growing cost of continually transferring large volumes of information to cloud platforms for processing. By analyzing information where it is created, manufacturers can identify defects, safety risks and process deviations immediately while keeping sensitive operational data within the factory environment.
The benefits extend well beyond technical performance. Frontline workers receive guidance while they are performing a task rather than after an event has been analyzed elsewhere. Quality issues can be identified before defective products move further through production. Maintenance teams can respond to emerging equipment problems before they become unplanned downtime. AI becomes an active participant in operational workflows rather than simply another analytical tool.
“We’re finding that on-device AI, whether it’s deep learning on a vision controller or AI agents on a mobile computer, gives frontline workers a solution without latency,” Pottel explains. “For leaders it means better cost control and more security because the data doesn’t have to leave the factory. Risks, defects and anomalies are picked up and acted on there and then, downtime is avoided and throughput is protected.”
This reflects a broader shift in manufacturing architecture. Competitive advantage will depend less on where data is stored and more on where intelligence is applied. For many operational decisions, the right place is increasingly the point where the work itself is taking place.
From connected factories to intelligent factories
Connecting more machines and collecting more operational data will soon become the price of entry rather than a source of competitive advantage. The manufacturers that pull ahead will be those that transform information into operational awareness, enabling people, systems and AI to respond as conditions change rather than after the event.
That evolution is already underway. As AI continues to mature, the focus will shift from individual applications towards intelligent environments where frontline workers and digital systems continuously support one another. “Computer vision, deep learning and machine vision are now being joined by AI agents or digital workers that operate within ambient intelligent environments,” he says. “This combination of human workers and agents creates what we describe as augmented collective intelligence, where workers contribute domain expertise and AI scales those talents through decision support.”
Rather than relying on a single, centralized AI model, manufacturers are likely to build distributed networks of intelligent devices, sensors and software, each contributing to a real-time understanding of operations while sharing information across the wider business. The result is not autonomous manufacturing for its own sake, but factories that become progressively better at anticipating problems, supporting decisions and adapting to changing operating conditions.
Ultimately, the real significance of ambient intelligence lies not in the technology itself but in the operational awareness it creates. “At the technical level, it’s about turning raw, multi-sensor data into real-time operational context,” he concludes. “At the human level, it’s about building environments where frontline workers are empowered with the foresight and clarity they need to make work better every day.”
