Lessons from the frontline: This is what AI does in the industrial world

I’ve spent my career in industrial automation—leveraging digital technology to control complex processes in sectors including energy, chemicals, and manufacturing. What began as a largely manual field has transformed into a dynamic ecosystem powered by autonomous systems tackling intricate, high-stakes challenges. This evolution closely mirrors the growth of the computer industry itself, marked by rapid innovation, increasing complexity, and expanding capabilities.

The early days: When AI was still unknown

In the late 1990s, I had the privilege of working for one of the first industrial AI companies. One of the founders was an MIT professor with deep expertise in AI natural programming languages. At the time, “AI” was a term often dismissed as theoretical and impractical. Instead, it was branded as Expert Systems, narrowly focused software models trained on specific variables to solve niche problems, given there was no massive computing infrastructure required or available at that stage.

These early systems were elegant in their simplicity: relying on cause-effect logic rather than an understanding of the entire process. They utilized an “expert system” approach based on rules and neural nets compared to generative AI, emphasized computational efficiency and precision over massive training and big data.

Back then, even with the rise of PCs, hardware lagged software capabilities. Training an AI model could take days—something that modern hardware now accomplishes in milliseconds.

Safety first: Why industrial AI adoption is complex?

In industrial sectors—such as refineries, power generation, steel, and cement—safety is paramount. The high volatility of fuels, extreme temperatures, and exothermic processes create environments with a low risk tolerance, so technologies must be thoroughly vetted and proven before being deployed. This risk-averse culture within many organizations further slows the adoption of new technologies like AI, especially in critical operations.

Automation in these industries generally consists of three layers:

1. Sensors and actuators (the physical interface)
2. Edge computing and control systems
3. Cloud-based analytics

If you liken this to how a car works:

• Sensors = steering and wheels
• Edge control = dashboard, cruise control
• Analytics = GPS and cloud diagnostics

To draw another analogy: In aviation, the most critical parts of the flight are taking off and landing which constitutes only a small part of the overall journey. The pilot controls the plane, and the rest is auto piloted. Similarly, in plant operations, startup and shutdown require human oversight and control, but the majority is run by automation.

Can we eliminate operators or pilots entirely? Technically yes—but in high-stakes environments involving human lives and billions in assets, society still demands 100% confidence, not probabilistic “advice” with 90% certainty.

AI is excellent at providing data-driven, intelligent insights, but still requires human decision-makers in critical situations—much like in medicine, where AI can suggest diagnoses, but final decisions rest with doctors.

Some practical industrial AI applications are already proving their immense value in targeted use cases:

• Steel: Optimize coke injection rates, enhance product quality, and predict asset failures
• Cement: Improve kiln efficiency and reduce energy use
• Upstream Oil & Gas: Methane leak detection aided by visual AI
• Refineries: Predict and prevent crude unit flooding by analyzing pressure and temperature patterns.
• Petrochemical: Reduce flaring and shutdowns by adapting to ambient conditions
• Soft Sensing: Replace expensive sensors by using AI to infer measurements from process data
• Polymer Production: Minimize off-spec product during transitions
• Biogas: Optimize feedstock blending and control for peak biogas yield
• Cybersecurity: Defend against increasingly sophisticated AI-enabled threats
• Microgrid Management: Forecast renewable generation, manage storage, and reduce emissions
• Asset Predictive Analytics: Predict industrial equipment failure to minimize unscheduled downtime
• Safety Analytics: From analyzing hazardous operations to implementing safety procedures

Before AI, industrial automation was already highly advanced:

• Advanced Process Control (APC): Often considered a precursor to AI, APC uses empirical models to control processes.
• Real-Time Optimization (RTO): Balances competing goals like efficiency, safety, and maintenance costs, helping users “do more with less” by analyzing trade-offs and constraints.
• Digital Twins: High-fidelity models based on first principles and mathematical representations of chemical processes allow for predictive control and optimization.

AI can build upon these foundations by automating model creation and learning from complex, multivariable data environments that traditional linear models struggle with.

AI and digitalization as a bridge to sustainability

The fastest way to decarbonize would be to cease industrial production entirely, but this is unrealistic—global demand for affordable energy, materials, and plastics remains high. Thus, the energy transition is about balance: ensuring progress while minimizing environmental impact.

Energy companies now face a dual mandate: achieve profitability while meeting sustainability goals. While renewables, green hydrogen, and carbon capture show promise, their ROI can be uncertain without regulatory support or infrastructure. That’s where AI and Digitalization can play a key role in improving efficiency, reducing waste, and supporting electrification without heavy capital investments.

Final thoughts

Industrial artificial intelligence is not a silver bullet, nor a replacement for skilled operators or engineers. But as part of a broader digital transformation strategy, it has immense potential. From improving safety and reducing emissions to optimizing production and enabling new forms of automation, AI is helping the industrial sector chart a more intelligent, efficient, and sustainable future.

The journey toward operational excellence, and digital innovation is just beginning—and AI will be a vital co-pilot along the way.

Learn more:Powering sustainable AI in the United States | Schneider Electric

Tags: Artificial Intelligence, cement decarbonization, digitalization, Safety, steel industry