AI in industrial automation: Getting the balance right
Udgivet kl. 21. maj 2026 i AI
Few technologies in recent memory have generated as much debate as artificial intelligence. Some predict it will transform industrial manufacturing almost overnight. Others dismiss it as overhyped and underdelivering. Neither view is particularly useful, and both can lead companies to make poor decisions.
The manufacturers extracting real value from AI are the ones who started with a deceptively simple question: what, precisely, are we expecting this technology to do?
People first, technology second
At OMRON, our guiding principle is that technology should serve people, improve lives, and contribute to society. That is not just a philosophical position. It has direct practical implications for how AI should be deployed on the factory floor.
Modern manufacturing is shaped by skilled people: operators, engineers, and managers who carry years of process knowledge that no dataset can fully capture. The role of AI is not to replace that knowledge. It is to make it more effective.
Consider a complex production line, where dozens or sometimes hundreds of variables are in motion simultaneously: temperatures, pressures, machine states, supply conditions, quality indicators. A human operator cannot monitor all of them at once. AI excels precisely at this kind of continuous, multi-variable monitoring. It does not get tired or distracted, and it can detect patterns in data that would take a person hours or days to identify. But when AI detects an anomaly, the question of what happens next is where clarity of thinking matters most.
Finding the right level of autonomy
One of the questions we are most often asked is where AI should be applied first. Our answer is usually the same: start where the value is clearest and where the limits are well understood.
Predictive maintenance is a good example. AI can analyse sensor data to identify changes in machine behaviour and raise a signal before a breakdown occurs. For narrow, repeatable, low-risk actions, AI systems are already operating autonomously and doing so effectively. But routine automated responses are fundamentally different from higher-stakes, contextual decisions that carry business consequence — those involving accountability, regulatory considerations, and the kind of situational judgement that comes from knowing an operation in its full context.
Our approach, developed through work at our own factories and with manufacturers across Europe, follows a clear progression: AI first analyses and detects, then begins recommending, and only earns the right to act autonomously in areas where operators have built genuine confidence in it. If a model changes, that trust must be re-established. It is also increasingly aligned with regulatory expectations. The EU AI Act establishes risk-based requirements for AI in industrial settings, including meaningful human oversight in higher-risk applications. Exactly the question manufacturers should be asking regardless.
Avoiding the traps
The overestimation trap is well documented. AI deployments fail for predictable reasons: data quality is not there, organisational readiness is lacking, or the problem definition was not sharp enough to begin with. But there is another failure mode we see repeatedly in the field: treating AI as a point solution. A company solves one specific problem well, but builds it in a way that cannot be replicated elsewhere. Scaling then becomes prohibitively complex. The companies that succeed ask from day one: how do we build this so it works across more lines, more factories, more products?
The underestimation trap is equally real. Companies that dismiss AI as relevant only for large enterprises with vast data science teams risk falling behind competitors quietly extracting genuine value from targeted applications. AI tools are becoming more accessible. The barrier to entry is lower than it was even three years ago. For European manufacturers already under pressure from energy costs, labour availability, and global competition, the cost of inaction is rising.
The path between these two traps is clarity: about the problem, the data, the people involved, and what a realistic outcome looks like. With the patience to start focused, prove value, and expand from a position of confidence rather than urgency.
Reasons to be optimistic
Labour demographics across Europe are creating real gaps in the industrial workforce. Operational complexity is increasing. The pressure to reduce energy consumption is not going away. AI, applied thoughtfully, can help manufacturers address all of these challenges, ensuring the people who are there can do more, decide better, and operate with greater confidence.
A 2025 study of 216 senior manufacturing leaders found that 75% expect AI to be a top-three contributor to operating margins by 2026. But the same study found that only 21% of organisations currently have full AI readiness. This is a substantial gap between ambition and the foundational work still required.
That gap is a reason for discipline. The manufacturers succeeding with AI are treating it as a practical question: where does this tool make our people more effective, and how do we implement it in a way that builds capability rather than dependency? Whatever their size or sector, that is the right question for every European manufacturer to be asking today.
