We all know that Operational Technology (OT) systems are critical for running the key services that make our society function, keep us out of the caves and from potentially eating each other.

In sectors such as energy, factories, food services, manufacturing and transportation, these all require stable and secure system function to make our physical world operate smoothly. As part of broader digital transformations, the use of Artificial Intelligence (AI) as a key tool – whether we like it or not – is becoming a bolt on many OT systems. The potential benefits are sold to us as being game changers, but there are risks present and these need to be understood prior to adoption.

In this article, I have tried to explore both sides of the argument and provide a perspective on why AI should and shouldn’t be used in OT. Hopefully this will provide some guidance on what is becoming a complex landscape.

So let’s be optimistic and start with the positives!

Why should AI be used in Operational Technology?
The potential that AI has to bring benefits to businesses of all types is undeniable, particularly when aligning with goals of Industry 4.0, the emerging fourth industrial revolution concerning the digitalisation of the manufacturing industry technologies and processes.

1. Predictive Maintenance and Reduced Downtime
   Predictive maintenance and condition monitoring has always provided great data sets which can be used to correlate maintenance related issues and potentially gain a deeper understanding of assets. Even with correlation not being causation, the potential to harness this data to get trends and make improvements has now become standard process.

   Using AI-powered predictive maintenance as an extension to these practices allows for the potential to further detect anomalies, predict equipment failures, and schedule maintenance before issues escalate with more accuracy across large data sets.

   An industry example of this is from Siemens, who use Generative AI to provide insights as part of their Industrial Copilot, a generative AI-based assistant that promises to assist with the development of code for PLCs, amongst many other use cases.

2. Enhanced Operational Efficiency
   AI can enable process optimisation through the use of algorithms and analysis of data from assets. Being able to detect inefficiencies and offer recommendations to adjust these can increase productivity and outputs, whilst reducing overheads such as energy usage. This has the potential to lower operational costs and help industries remain competitive.
3. Improved Worker Safety
   This really should be number one and a no-brainer when it comes to OT systems. There is potential for AI to enhance safety in hazardous OT environments by monitoring conditions and predicting risks.

   As an example, in oil and gas operations, being able to analyse sensor data for failures in equipment and provide alerts prior to an incident can reduce risks and human error and ultimately provide greater protection to workers.

4. Advanced Cybersecurity Threat Detection
   With OT systems being increasingly connected to cloud and enterprise environments, we are seeing an increase in cyber attacks. Traditional systems that may be on premise are not generally able to leverage the large data sets that provide context around the threat landscape.

   The integration of AI into OT security platforms to provide automated threat response and behavioural analysis can aid in safeguarding critical infrastructure from sophisticated attacks.

5. Digital Twins and Process Optimisation
   By using digital twins, which are effectively digital copies of physical assets, OT processes can be simulated and optimised. Through the prediction of outcomes and scenarios this can accelerate innovation, reduce costs and support sustainability by minimising resource waste and aligning with Industry 5.0’s human-centered goals.

Why Shouldn’t AI Be Used in Operational Technology?

There are a lot of promises and up-selling done using AI by just about every vendor on the market, but putting the hype aside for a second, there are risks associated with adopting AI across such a broad range of OT functions. Looking at the positive use cases as mentioned above, what are the impacts (if any) of those fails, based on potentially inaccurate or just plain wrong analysis? This is why the adoption of AI into OT environments requires careful thought and planning.

1. Heightened Cyber Security Vulnerabilities
   The dichotomy of enhancing OT cyber security detection and response can also induce new risks through connectivity. AI systems are prime targets for advanced adversaries who may seek to poison data, exploit vulnerabilities and bypass security measures. The sharing of sensitive OT information, designs, diagrams, network models etc. with potentially public models can add a new threat vector which can be difficult to unscramble.
2. Lack of Explainability and Trust
   AI systems typically operate as black boxes, meaning that there’s no traceability of the decision-making process. Within OT environments, any decision can impact safety and these need to be auditable and transparent. The use of AI can erode trust. Examples such as halting a manufacturing process or stopping production without clear explanation can erode trust in the safety of the systems, particularly when following frameworks such as IEC 62443, which emphasise transparency. AI struggles to provide this, which complicates compliance.
3. Bringing AI into Operational Technology
   Introducing AI into Operational Technology can be like embarking on a monumental project, requiring lots of money and talent. It’s not just about updating hardware. Companies must hire smart data experts and OT specialists to design and support AI systems, which can be very costly. Small and medium sized businesses often struggle with these high costs, especially if they’re unsure about getting a good return on their investment. Moreover, integrating AI with OT systems such as Modbus and OPC UA can feel like tackling a complex puzzle, leading to delays and operational risks.
4. Risk of Dependency and Skill Loss
   When companies rely heavily on AI, OT workers might lose their touch with manual skills. If an unexpected crisis hits, like a cyber attack that disables AI, will the workers have the knowledge and expertise to manage without it?
5. Regulatory and Ethical Quagmires
   Using AI in OT requires navigating a maze of stringent regulations such as NERC CIP, NIST SP 800-82, and the EU’s DORA. These regulations demand thorough testing and validation, which AI systems may struggle to achieve because they are always evolving. Additionally, ethical questions about responsibility arise. If an AI-caused decision results in a safety issue, who takes the blame? These unanswered questions often make OT leaders cautious about adopting AI technology.

Conclusion: A Transformative Yet Cautious Path Forward

Yes, AI holds immense potential for Operational Technology environments, offering breakthroughs in efficiency, safety and innovation that are reshaping industries in 2025. Predictive maintenance, digital twins and cyber security enhancements are just the beginning of what AI can achieve in OT. However, the risks, such as safety compliance, risk reduction, cyber security vulnerabilities, legacy system challenges and regulatory hurdles, demand a more cautious approach. By starting small, prioritising security and fostering human-AI collaboration, organisations can harness AI’s potential whilst also safeguarding critical operations.

As OT continues to evolve, the question isn’t whether AI should be used, but how it should be implemented. Anchoram Consulting feels that this approach is aligned with known sources of authority, such as key technical standards and advice.

With careful planning and strategic investment, AI can be a catalyst for progress in OT, driving industries toward a smarter, safer and more sustainable future.

For more information on how Anchoram Consulting can assist with your journey into a modern, safe and resilient OT environment, contact us today!