What you’ll learn:
- As they navigate the challenges of an increasingly connected and complex marketplace, OT teams will need AI to help them reach beyond what they can do today.
- Time spent filtering information noise undermines confidence in AI as an adviser.
- Many OT teams are finding their automation solutions provider is an invaluable resource for deploying fit-for-purpose, effective AI for operations.
Operational technology has come to a crossroads. Over the last few decades, process manufacturers have taken advantage of the decrease in cost and sharp increase in capability of sensing equipment to instrument and monitor more areas of the plant than ever before.
These new sensors are creating volumes of structured and unstructured data far faster than OT personnel can sift through it and triage it manually.
OT teams also have discovered the value of data.
For years, plant personnel wondered if they were collecting too much data and how they could manage it, then along came the rise of AI tools, demonstrating new ways to harness copious amounts of data and turn it into actionable information.
Today, a wide variety of AI software applications thrive on large volumes of data.
See also: Crystal Ball 2026 Series
However, for a group like OT, centered around uptime, risk mitigation, determinism, proportional-integral-derivative loops, control strategies, and first principles methodologies—the intersection of big data and modern AI creates new challenges.
Generative AI’s rise has shown how large models can synthesize and summarize information; yet without constraints tied to thermodynamics, equipment design limits, mass and energy balances, loop parameters, and safety interlocks, an AI system can confidently propose conditions or interventions that violate process reality.
In the OT realm, a plausible‑sounding but incorrect recommendation may erode operator trust, delay decisive action, or introduce risk in real-time, critical, low-latency environments.
As they navigate the challenges of an increasingly connected and complex marketplace, OT teams will need AI to help them reach beyond what they can do today, providing more visibility and predictive capability, while increasing opportunities for innovation.
In addition, they need to do so in a way that combines computational fact with computational intelligence, so the resulting information is useful and actionable for the operator residing between the AI and production systems.
Unbounded AI is inaccurate
Consider an example of an operator shift change handoff. A practical use for AI in such a scenario enables the new shift operator to simply ask an AI adviser, “What should I know about what happened on the previous shift?”
That question, however, is extremely open-ended. A well-tuned AI tool might respond with information the operator needs, citing production rates versus targets, a small set of rationalized alarms, a distillation column approaching its energy efficiency limit, or a quality trend edging towards an off-spec boundary.
Industrial AI is built and supported by decades of OT and industry-specific expertise and delivered via an enterprise operations platform built on seamless data mobility.
However, if the AI is working with an untuned enterprise-wide data pool, it might clutter the response by inserting irrelevant personnel notes about a coworker on the previous shift clocking in late or information about transient IT outages.
Time spent filtering information noise undermines confidence in AI as an adviser. Reliability hinges on scoped and contextual relevance, rather than linguistic fluency.
State of Initiative survey: Mixed signals on AI in investment plans, production systems
Some of the challenges with hallucination and misinterpretation from AI can be controlled by context and continual, extensive training of the AI model. Yet, most teams implementing AI software are investing for productivity. If those same teams must also spend hours continually tuning the AI and carefully evaluating its every output, the productivity gains will be minimal.
The solution is industrial AI—persona-based, first principles-driven AI software, built and supported by decades of OT and industry-specific expertise and delivered via an enterprise operations platform built on seamless data mobility.
Fit-for-purpose artificial intelligence
Industrial AI relies on immutable first principles constraints, designed around detailed physics, chemistry, equipment designs and control constraints that define feasible operating states.
By bounding the solution space, the system provides recommendations that are relevant, possible, and within budget to provide guided insights based on meaningful data.
Effective industrial AI must also be persona driven. The software needs to understand individual users’ unique responsibilities. An operator, a reliability engineer, and a process engineer benefit from different recommendations and information.
Supporting industrial AI
Because industrial AI is built on a foundation of first principles and persona-based knowledge, it requires an equally robust data source for support. Industrial AI models can consume and deliver value from raw, unstructured data, but their impact improves exponentially when driven by contextualized data from a wide variety of applications.
To meet this need, many organizations are building their automation foundation on an enterprise operations platform model—an infrastructure designed for seamless data mobility via a data fabric that moves contextualized data effortlessly from the intelligent field, through the industrial edge, and into the cloud.
Industrial AI relies on immutable first principles constraints, designed around detailed physics, chemistry, equipment designs and control constraints that define feasible operating states.
As a result, many OT teams are finding their automation solutions provider is an invaluable resource for deploying fit-for-purpose, effective AI for operations.
The most advanced automation solutions providers not only offer the infrastructure to deliver seamless data mobility, they develop and deploy industrial AI tools integrated into those frameworks and built on decades of first principles and industry-specific knowledge.
Such a strategy leads to more seamless, scalable deployment of AI with more reliable results than bolt-on solutions.
As AI solutions continue to evolve, there has never been a better time to consider an OT modernization strategy built on seamless integration driven by a comprehensive data fabric.
A new industrial era is on the horizon, and those that prepare today will capture the competitive advantage that will drive success for decades to come.
About the Author
Brian LaMothe
Brian LaMothe is VP of applied research and emerging technologies for Emerson’s process systems and solutions business. He has more than 25 years of experience in distributed control systems and supervisory control and data acquisition environments, with a career spanning product development, project management, operations, IT, digitalization, and business process improvement.