Monday 18 May, 2026
“The future doesn’t belong to companies that have the fastest network or the best technology. It belongs to partnerships that speak the same language. The language of measurable value.” Those words from Priya Kurien, Research Director at the IBM Institute for Business Value, at the Manufacturing and Production Summit at MWC26 Barcelona underscored a central point in the debate about how to realise industrial transformation at scale: manufacturers are not looking for isolated technologies, but for partners that can solve measurable operational problems. This is because manufacturing has moved into a more demanding phase. Now, the critical issue is whether technologies such as AI, private networks, automation and digital twins can be combined to solve operational problems at scale. Across the summit, the most serious contributions pointed in the same direction. Manufacturing does not need more siloed pilots. It needs integrated systems that improve resilience, productivity and control.
Modernisation is critical because pressure on manufacturers is growing from several directions at once. Supply chains remain fragile. Labour shortages are becoming structural. Product cycles are shortening. Cyber risk is widening as more operational environments become connected. AI is also raising expectations that industrial systems should become more responsive, more predictive and more autonomous. Yet that promise only becomes useful when it is tied to the real constraints of production. As Priya Kurien put it, “What manufacturing buys is not connectivity. They buy outcomes.”
Networks, edge computing and AI are not meaningful to manufacturers as abstract capabilities. They matter when they reduce downtime, improve quality, help plants respond faster, or make scarce expertise go further. Too often, suppliers still present infrastructure as the product, when manufacturers are looking for business outcomes.
The pilot phase is ending, but scale is still difficult
As Tunc Yorulmaz, Global Head of Networks at Accenture explained, the current juncture in industrial transformation is the point where “we pass through POC cycle” and where “everyone is asking what is real”. The immediate question now is whether a system can be shown “in a production system” and in “a real implementation.” This is a reflection that the market is becoming less interested in demonstration and more interested in deployment.
Industrial transformation needs joined-up systems, not siloed offers
One of the recurring snags facing the industry is that the provider ecosystem still mirrors the internal fragmentation of large manufacturers. Strategy, IT, operational technology and plant operations often sit in separate silos. Many solution providers operate this way. That makes transformation harder to scale, because no single offer addresses the whole operating environment.
John Robinson, Founder and Creator of the Quorum Principle argued that “there’s a systemic problem in global manufacturing” and that “the solution provider ecosystem is an exact mirror image” of the same fragmented areas of expertise inside manufacturers. His point was not that specialist vendors lack value. It’s that manufacturers are often forced to assemble a patchwork of disconnected solutions and then make them work together themselves. That is one reason so many pilots fail to expand into broader transformation.
This challenge also explains why partnerships are a recurring subject of debate surrounding industrial transformation. Yet partnerships and collaborative efforts depend on alignment. As Priya Kurien argued, “partnerships don’t fail because of technology. Partnerships fail when incentives, ownership, and success metrics are not aligned to the same outcome.” That is a much more practical test. Integration will remain partial unless the network provider, software company, integrator and manufacturer all work to the same business result.
This is why the commercial model matters as much as the technical one. Tunc Yorulmaz warned that enterprises increasingly want faster engagement, new commercial structures and “an outcome-based approach” rather than traditional charging logic. That fits what manufacturers are asking for. They are trying to improve throughput, reduce downtime, shorten lead times and cope with labour shortages. They are not trying to buy 5G as an isolated product.
AI is becoming useful, but trust and control still decide adoption
AI deployments have led to gains in visual inspection, anomaly detection, field service and robotics support. Diego Romeres, Team Leader and Senior Principal Research Scientist at Mitsubishi Electric Research Laboratories, said AI is already beyond hype when it improves “the KPI that’s trying to optimise” and cited “high speed visual inspection”, anomaly detection and machine status support on the shop floor as examples of real value.
Yet industry still distinguishes between useful AI and fully trusted autonomy. When AI supports decisions, organisations can adopt it more quickly. When it starts acting inside physical control loops, questions of stability, verification and safety become much harder. As Romeres explained, trust in this context means predictable behaviour, verifiable performance and the ability to fail safely.
This caution underscored the reality that factories are not about to hand control to autonomous systems wholesale. It means narrow, high-value applications are becoming credible, but their expansion depends on stronger governance, better connectivity and more disciplined validation.
A clear example of how trust is reinforced came from the complementarity between Dassault Systèmes and Archetype AI. Dassault showed how physics-grounded virtual twins help engineers test and validate designs with confidence in the virtual world. On the other hand, Brandon Barbello of Archetype AI applied that same logic into the factory. He argued that “the physical world speaks in a different language. It’s a language of vibration and temperatures and pressures and visual spatial data that we don’t have on the web.” Together, these approaches suggest that manufacturers will trust AI more when its outputs can be tested, explained and checked against operational reality.
Security and resilience are critical to industrial transformation
Once industrial environments become more connected, cybersecurity can no longer sit outside the transformation story. It becomes part of the operating model itself. Digital risk can come from several angles: expanded attack surfaces, legacy OT environments, supply chain exposure and data sovereignty.
Amir Stephenson, Director of Networked Electromagnetic Systems for Lockheed Martin, broadened the discussion beyond cyber risk alone. He argued that secure industrial transformation depends on maintaining “persistent, secure information flow across a hybrid network” through the deliberate pairing of commercial and non-commercial technologies. In other words, resilience comes not just from blocking threats, but from designing networks that can operate securely across complex, mixed environments.
David O’Byrne, VP Business Development at Druid Software described the security situation starkly: industrial transformation has “created the largest attack surface in the history of attack surfaces”. Yet Steve Shaw, 5G Solutions Marketing Lead from Palo Alto Networks argued that security teams must stop acting as “the group of no” and instead help shape projects from the start. Those remarks matter because they shift security away from compliance language and into operational language. In manufacturing, a cyber incident is not just an IT event. It can mean disruption to production, exposure of intellectual property, or loss of control over essential systems.
Sovereignty is likely a long-term issue
That leads to a broader point the industry has to contend with about sovereignty. The debate here is not reductive digital nationalism. It’s whether some industrial workloads, especially those involving sensitive engineering data, simulation or strategic infrastructure, require stronger control over where data sits and who governs access to it. In that sense, resilience now means more than redundancy. It includes trust in infrastructure, in data handling and in the continuity of operations.
These issues demonstrate how manufacturing is evolving and point to a more disciplined stage of digital change. The necessary technologies involved are becoming more sophisticated to meet long-term commercial demand. But progress now depends less on showcasing innovation and more on integrating it. The next gains will come from connecting AI, networks, automation and cybersecurity into one operational model that can stand up in production. That is a harder task than running pilots. It is also the one that now matters most.
