There has been an AI surge in manufacturing. You are surely aware of this.

Recent advances in artificial intelligence have yielded remarkable demonstrations in speech, language, code generation, and content creation. These developments have generated enormous interest and growing demand for AI/ML applications.

Unsurprisingly, these advances have been eagerly watched within manufacturing by engineers, factory operators, and technology leaders to explore how AI can improve quality, reduce rework, and unlock incremental throughput.

Yet despite the excitement and significant investments, many organizations still struggle to translate AI demonstrations into business value and are discovering that deploying these systems within production environments is far more complex than adding a chat interface or connecting a language model to existing data sources. Perhaps you are such an organization.

This challenge should not be surprising. Industry has encountered a similar cycle before.

Lessons from the first AI wave

More than a decade ago, the first major wave of industrial data science and machine learning (DS/ML) fueled Industry 4.0 initiatives. Early expectations were substantial. Major investments were made in data and connectivity platforms to build the data layer first. Descriptive analytics, exploratory data analysis, and predictive analytics projects proliferated.

Despite a wealth of opportunities, when DS/ML was first applied to industrial datasets, a large percentage of projects failed to achieve operational value. A key reason was that the first generation of algorithms were designed primarily for probabilistic internet behavior, consumer interaction, advertising, and recommendation systems, not for deterministic industrial environments requiring safety, physical validity, and repeatable results.

The current wave of AI projects may follow a similar trajectory, and early signs are already emerging. In industrial manufacturing, GenAI, copilots, foundation models, and industrial agents are simplifying aspects of automation and expanding access to advanced capabilities. At the same time, their adoption mirrors early ML, but at a much larger scale.

While modern AI excels in language, summarization, and generating plausible answers, factories require real-time grounding, process physics, safety guarantees, and regulatory compliance.

Manufacturers increasingly ask, “We have all the data; now tell us what to do with it.”

Automation intelligence bridging the gap

Understanding how AI can be successfully applied requires understanding the technology itself, its relationship to the broader field of data science, and the lessons learned from the first wave of Industry 4.0. Drawing upon these lessons formed the foundation for automation intelligence, a technically grounded framework that, when combined with current AI tools, increases success when applied to challenging industrial problems.

Contemporary advances in AI are largely driven by large language models (LLMs). At their core, LLMs learn statistical patterns in language to predict what comes next in a sequence. Combined with advances in computing power, they can generate highly coherent and contextually relevant responses. Imagine asking a question and having the entire text of the internet available to formulate an answer. It is therefore not surprising that these methods can produce convincing responses.

In many industrial settings, however, these systems behave more like advanced search and synthesis engines unless they are properly grounded. Practitioners should understand these limitations to position applications of AI appropriately, with realistic expectations.

AI is prone to hallucination, a common term to describe factually inaccurate or incorrect responses. This limitation should remind practitioners of the first wave of industrial DS/ML, where successful applications ultimately emerged by introducing engineering constraints, domain rules, and methods for tailoring algorithm outputs.

Get your subscription to Automation World's tri-weekly newsletter.

Despite significant advances in adaptability and throughput, the current generation of AI faces the same fundamental limitation: industrial outputs must satisfy constraints on accuracy, safety, and stability that are not native to AI architectures. Agentic workflows can help mitigate hallucinations and impose output constraints, but without process context and engineering rules they may still fall short of industrial requirements while introducing computational complexity.

Automation intelligence bridges this gap. By applying engineering-derived constraints to the inputs and outputs of AI, it allows the actions derived from AI output to be applied effectively to industrial systems today. This approach unlocks immediate value while positioning organizations to evolve current AI applications to the next frontier of vision-language-action (VLA) models.

Automation intelligence in practice

Consider a simple example: we ask an AI, “How fast is the car traveling?” The response will almost certainly be a speed estimate or a method for calculating speed and is unlikely to return an unrelated quantity. Even this output has value: it can reduce commissioning time, narrow root-cause investigations, and assist in training new operators.

However, a basic understanding of AI reveals that the answer is generated from learned language patterns rather than direct awareness of the physical system. Automation intelligence, while not part of current off-the-shelf AI techniques, can be combined with existing AI methods to provide key contextual constraints to improve reliability. Examples include:

1. Speed limits. A moving car typically operates within a range near the posted limit.
2. Distance to the vehicle ahead. Assuming the leading car is following the speed limit, changes in following distance constrain our likely speed.
3. Physical vehicle limits. The speed is bounded by the vehicle’s mechanical limits. 

These rules represent process context and provide constraints for AI. While output validation remains important, post-processing approaches such as agentic workflows act after the fact and come with increased computational burden. Automation intelligence instead applies engineering constraints to AI so it can act as a disciplined layer integrated with industrial control systems.

Unlocking industrial AI’s potential

Automation intelligence can accelerate AI adoption across industries including food and beverage, automotive and tire, semiconductor, oil & gas, consumer packaged goods, and pharmaceutical. Within these verticals common applications include discrete and continuous processes such as drying, chemical synthesis, assembly, extrusion, packaging, rolling/winding, purification, and mixing.

Although many of these processes have been optimized for decades, AI creates new opportunities for value. Automation intelligence accelerates the path to industrial value, while improving deployment success.