Manufacturers must separate GenAI’s promise from reality, confront cultural resistance, and stay focused on collective intelligence if they want their MES strategy to succeed in the AI era. Rick Franzosa was a keynote speaker at Critical Manufacturing’s MES & Industry 4.0 Summit in Porto, and that was his key takeaway.
The job of a manufacturing analyst is not to cheerlead. It is to apply healthy scepticism to every new wave of technological hype and ask what truly delivers value. Rick Franzosa, formerly a research vice president at Gartner and now a principal analyst at Gormally Franzosa Associates, has spent decades doing just that. For him, the GenAI conversation is moving dangerously fast, and large sections of the manufacturing world are simply not ready.
“Vendors are frustrated because manufacturers still expect a red button in the sky that solves all their problems,” Franzosa explains. “When the field engineers arrive and ask where the data is, what matters, and how it is structured, they are met with blank stares.”
That gap between expectation and reality sits at the heart of the current confusion surrounding MES and AI. The truth is not that GenAI is irrelevant or overhyped; instead, the groundwork remains critically incomplete. Franzosa is clear: the issue is not technology but mindset. Cultural resistance, cognitive entrenchment and broken data practices are the real obstacles.
AI needs people more than platforms
The Gartner Hype Cycle famously charts the arc from inflated expectations to disillusionment. GenAI, Franzosa notes, is hurtling down that curve, but it is not slowing down. Of the 24 AI technologies mapped by Gartner in late 2023, only four were expected to mature within three years. By 2025, the curve had already shifted again. Franzosa describes this acceleration as both exciting and alarming.
The challenge is not that GenAI does not work. The challenge is that organisations are implementing it without the foundational capability to support it. The myth of instant intelligence, where pressing a button yields answers, remains one of the most dangerous assumptions in digital manufacturing today.
This lack of readiness is especially evident in the workforce. According to Franzosa, “Even the most optimistic advocates of AI acknowledge that human decision-making will remain essential for decades,” he adds. “Rather than displacing people, GenAI shifts the focus towards how frontline workers engage with data, process and technology. The quality of outcomes depends entirely on how well human judgement and AI capability are integrated. Everyone says that people, process, and technology matter, but what AI teaches us is that order matters. It must be people, process, data, technology, in that order.”
This hierarchy is not just rhetorical. Franzosa references Tech-Clarity research, which shows that workforce development is now considered twice as critical to business success as digital transformation. Yet, the same research shows that workforce disruption tops the list of business risks. The contradiction is stark: manufacturers recognise the importance of people yet continue to underinvest in training, inclusion, and trust.
Psychological reactance, the tendency to resist anything that threatens perceived autonomy, is rampant. Franzosa argues that most frontline workers have no idea what AI means for their jobs. When told a new system is “good for the company,” many assume it is probably not good for them. The result is silent sabotage, cultural resistance and failed adoption.
Franzosa cites the pandemic as a significant example of this phenomenon. The language used around public health measures, phrases like “you must” or “you are required”, triggered a backlash in ways that a more inclusive, cooperative framing might have avoided. In manufacturing, the same dynamic plays out when transformation is imposed from above without adequate engagement from the people who will be affected by it.
Cognitive entrenchment is the real blocker
Technological resistance is often blamed on complexity or integration costs. Franzosa suggests it is more often due to cognitive entrenchment, the unconscious but deeply held belief that the current way of doing things is the only sensible one. Once a process is embedded in daily operations, even when it is flawed or outdated, it becomes psychologically protected.
Changing those processes requires a cultural shift that encourages rethinking and unlearning of existing practices. Franzosa uses a vivid metaphor to explain the depth of this challenge: “Cognitive entrenchment is like a poster on the wall that says ‘If you always do what you always did, you will always get what you always got.’ The problem is, the same people who hang the poster are often the first to resist change.”
His antidote is collective intelligence. Drawing inspiration from open science and medical research communities, he outlines three attributes of effective collective decision-making: diversity of expertise, decentralised authority, and the ability to aggregate and synthesise inputs across disciplines. “Technology helps,” he says, “but the real value comes from the dance across functions, behaviours and cultures.”
In manufacturing, this means dismantling functional silos and engaging people who ordinarily do not have a voice. It also means rejecting overly simplistic narratives that rely on a single source of truth. “I hate the phrase,” Franzosa adds. “It implies that all bias can be eliminated. It cannot. We must remain vigilant, reflective and humble.”
Franzosa refers to the Michael J. Fox Foundation’s Parkinson’s Progression Markers Initiative as a model of collective intelligence at scale: 42,000 online participants, 25 million data downloads, 1,200 scientific publications, and 40 industry partners. “That is what coordinated effort looks like,” he explains. “That is the power of data used to benefit a common goal.”
Data must be treated as infrastructure
For MES to become a real-time intelligent system, manufacturers need to fix the way they manage and contextualise data. Franzosa again points to a recent Tech-Clarity survey where top-performing manufacturers were 20 per cent more likely to manage all plant data consistently and integrate it with enterprise systems. They were also significantly more advanced in contextualising OT data for operational execution.
These organisations are not only treating data as an asset, but they are also treating it as a form of infrastructure. This involves having a clear governance framework, understanding the origin of data, identifying trustworthy data, and aligning that data with the business and operational context.
“MES does not fail because the software does not work,” Franzosa explains. “It fails because the data is inconsistent, irrelevant, or siloed. That is not a technology problem. It is a leadership problem.”
This data challenge extends beyond internal systems. As manufacturers deploy AI tools that rely on external models, such as large language models (LLMs) or third-party analytics engines, ensuring data fidelity becomes increasingly essential. Synthetic data, while promising in specific domains like simulation, cannot replace ground truth. Franzosa warns that reliance on artificial datasets introduces new risks, including false correlations, model bias, and operational drift.
Agentic AI demands more than optimism
The current fascination with agentic AI, systems capable of autonomous planning, reasoning and action, is understandable. Franzosa recognises the potential but warns against magical thinking. The robotic process automation (RPA) boom offers a cautionary tale: thousands of bots deployed with little documentation, accountability or maintainability. GenAI risks repeating that mistake unless controls are embedded from the start.
“RPA looked brilliant until nobody could remember what the bots were doing,” he notes. “GenAI will be no different if we are not careful. Without structure, we will automate confusion.” This includes understanding the difference between augmented and autonomous intelligence. Most successful applications today are still based on human-in-the-loop systems, augmenting decisions rather than making them. Trust, transparency, and explainability remain essential.
Franzosa also challenges the assumption that agentic AI will replace existing MES platforms. “The better question is how MES and AI can evolve together,” he says. “That means building interfaces, APIs, governance layers and feedback mechanisms that make sense in the context of manufacturing execution, not just retrofitting enterprise AI into factory systems.”
In terms of deployment, Franzosa emphasises the need for transparency. He points to new frameworks, such as visual agent tracing and action trails, that help manufacturers monitor how decisions are made. These tools are vital in regulated industries and for building trust within the workforce.
MES will evolve, but humans must lead
The final message from Franzosa is not pessimistic. He is no AI nihilist. He believes the top-performing manufacturers are already proving that AI can support smarter decisions and continuous improvement. But they succeed not because they chase hype. They succeed because they align AI with process discipline, data management and workforce inclusion.
He also believes that human intelligence, creative, imaginative, and interpretative, will remain central. MES may become smarter, more adaptive and more autonomous. But it will not become the brain. “You are the brain,” he concludes. “AI is a tool. MES is a system. The future will be defined by the humans who bring them together.”
Franzosa reminds manufacturers that continuous improvement is not a new concept. What has changed is the volume and velocity of data, as well as the tools available to make sense of it. Success will not be measured by the number of models deployed but by the ability to harness those models to make better decisions. That is not an AI problem. That is a leadership opportunity.
