I recently sat down with a group of undergraduate students planning a career transition into OT cybersecurity. Their first question was not about certifications or salary. It was whether AI was going to make the whole field irrelevant before they finished learning it.

I recently sat down with a group of bright undergraduate students who were planning their career transition into OT cybersecurity. Their first question was not about certifications, or salary, or which tools to learn. It was this:

“Sir, by the time we finish our specialisation and land a job, will AI already be doing it?”

Fair question. Every news cycle has an article about AI replacing knowledge workers. Cybersecurity is firmly on the list. So here is what I told them, honestly.

The short answer: no, but your job will change

AI will not take your OT cybersecurity job in the next ten years. But the OT cybersecurity job you start in 2026 will not be the job you hold in 2030. If you refuse to evolve, AI will not replace you. A human using AI will.

Let’s look at what is actually happening.

What AI does well in OT security

• Anomaly detection at scale.Modern passive monitoring tools already use ML to baseline OT protocols and flag deviations. This is genuinely better than rule-based detection, because the rule set in any real plant is too large for humans to maintain.
• Log triage.An L1 analyst spends most of their day looking at alerts that turn out to be nothing. AI can shrink that pile to something a human can actually reason about.
• Documentation and report drafting.This used to take days. It now takes a few hours, and the humans spend the saved time on the parts of the report that matter, the judgements.
• Pattern recognition across incidents.Humans remember the last five incidents. AI can remember all of them, across organisations, if we let it.

What AI does badly, and will keep doing badly, in OT

• Understanding physical consequence.An AI can tell you a valve command is anomalous. It cannot tell you that this particular valve sits between a hot stream and a cold stream and that closing it at this moment will cause a pressure spike that trips the whole unit. That is process knowledge, and process knowledge lives in the head of a chemical engineer who has been at the plant for 12 years.
• Stakeholder negotiation.Most OT security work is getting the plant manager, the IT director, and the vendor to agree on something. AI does not sit in meetings. Humans do.
• Liability.If an AI-driven control action causes a safety incident, no AI vendor is taking the fall. A human has to sign off. That human has to understand what they are signing.
• Novel context.When your plant does something unusual (a commissioning, a cyber-physical incident response, a new regulatory inspection), the AI has no training data for your specific situation. A human does.

What I told the students to actually do

Three concrete things:

1. Get fluent in the domain, not just the tools.Read about chemistry if you will work in chemical plants. Read about power engineering if you will work in utilities. The AI will always know the tools; your edge is knowing the plant.
2. Use AI as an intern, not a crutch.Let it draft your reports, summarise your logs, generate your first-pass threat models. Then edit with judgement. Students who refuse to use AI will be slower than those who use it. Students who accept AI output without editing will be wrong a lot.
3. Develop the skills AI is worst at.Communication. Negotiation. Physical-world reasoning. Ethics. These are human skills, and in OT they matter more than anywhere else in cybersecurity.

The real risk

The risk is not that AI eliminates OT cybersecurity jobs. The risk is that it raises the baseline productivity so much that teams become smaller. The jobs that exist will expect more output per person. That is not unique to our field. It is the story of every professional discipline since the spreadsheet.

Prepare for that world. It is already arriving.

One last thing I told them

“Ten years from now, you will be the senior person in the room with the AI at your disposal. Do you want to be the one who trusts it blindly, or the one who knows enough to challenge it when it is wrong? Pick now. Start studying accordingly.”

They went quiet for a bit. Then they started asking better questions. That is always a good sign.

RelyBlue’s OT security training includes modules on AI-assisted analysis, threat hunting, and how to critically evaluate AI output in safety-critical environments. See the training catalog.