For years, industrial automation has been moving toward smarter machines, better data, and tighter IT/OT integration. In 2026, that shift is no longer theoretical. AI is now showing up inside real engineering workflows, from code generation and HMI support to diagnostics, maintenance and production optimization. Vendors, standards bodies and government-backed initiatives are all treating industrial AI as a serious operational topic, not a side experiment.
The important point for engineers is this: AI is not replacing PLC programming. It is changing how PLC programming gets done. The engineers who do well over the next few years will not be the ones who blindly trust AI, and not the ones who ignore it either. They will be the ones who know where AI helps, where it fails, and how to use it without putting reliability, safety or cybersecurity at risk. That last point matters because manufacturing remained the most attacked industry in 2025, and industrial systems still face significant exposure from phishing, malicious scripts and internet-borne threats.
Why AI is becoming a real industrial automation topic
The pressure is coming from multiple directions at once. Manufacturers are still investing in physical automation, sensors, connectivity and analytics because those are the building blocks for more advanced AI-driven operations. Deloitte reports that 41% of manufacturers say factory automation hardware is a top priority over the next 24 months, with active sensors and vision systems also ranking highly. PwC’s 2026 outlook says the median share of highly automated processes is expected to more than double by 2030, from 18% to 50%.
The institutional signal is just as strong. In late 2025, NIST announced a $20 million investment to establish two centers focused on AI for manufacturing and critical infrastructure cybersecurity. Around the same time, ISA published a position paper on industrial AI that highlighted practical use cases like predictive maintenance, robotics, digital twins and real-time optimization, while also stressing the need for secure, standards-aligned adoption.
Then you look at the vendor layer. Siemens has made AI a central part of its current automation message. In 2025 it introduced AI agents for industrial automation and said its Engineering Copilot for TIA Portal can generate automation code from natural-language prompts. By SPS 2025, Siemens was describing Engineering Copilot TIA as a tool that could autonomously execute engineering tasks such as code generation, testing support, visualization development and workflow automation. Separately, TIA Portal V21 added a new export format for Git-based version control and CI-style workflows, which matters because AI-assisted engineering only becomes trustworthy when changes are traceable and reviewable.
Where AI is actually useful for PLC engineers today
This is where the hype needs to be separated from the real value.
The first strong use case is code scaffolding. AI is good at generating repetitive boilerplate, such as state-machine skeletons, alarm structures, UDT concepts, naming conventions, string handling helpers and draft SCL blocks. That does not mean the first answer is production-ready. It means the engineer starts from a faster draft instead of a blank screen. Siemens’ own positioning around Engineering Copilot TIA reflects that exact pattern: reduce repetitive engineering work so the engineer can focus on higher-value decisions.
The second use case is HMI and documentation support. AI can help generate alarm texts, operator messages, screen descriptions, tag explanations, commissioning checklists and troubleshooting guides. For engineers who build both PLC and HMI projects, this is low-risk, high-return work. The quality still needs review, but the time savings can be real.
The third use case is troubleshooting and log interpretation. AI is good at summarizing long fault histories, looking for patterns in machine events, turning messy notes into structured findings, and suggesting likely causes to investigate. ISA’s recent industrial AI paper highlights data analysis, maintenance and optimization as areas where AI is already meaningful in industrial contexts.
The fourth use case is maintenance and operations support. Siemens says its operations and maintenance copilots are being built to help operators and maintenance teams query machine data, receive guidance and move faster through troubleshooting. That fits a broader industry direction where AI is used to shorten the path from raw machine data to action.
Where AI should not be trusted on its own
Here is the hard truth: AI can write convincing nonsense.
That matters a lot in industrial automation because a PLC project is not an essay. A bad answer can stop a line, damage equipment or create a dangerous state. So there are some clear boundaries.
AI should not be trusted to make final decisions on safety logic, machine interlocks, emergency behavior, motion control, or process sequences with real plant risk without full engineering review, testing and validation. ISA’s industrial AI guidance explicitly frames adoption as risk-informed and standards-driven, with human safety, reliability, data quality, explainability and information protection all treated as core factors.
It also should not be allowed to make undocumented online changes to a live system. If AI is going to touch engineering, then version control, traceability, testing and rollback are non-negotiable. That is one reason Siemens’ move toward Git-compatible exports in TIA Portal V21 is more important than it might look at first glance: it supports the review discipline AI-generated changes need.
A safe workflow for using AI in PLC programming
The right way to use AI in industrial automation is not “ask for code and download to the CPU.”
The right way is this:
Start with a tight engineering brief. Define the machine behavior, inputs, outputs, fail states, mode logic, alarms and assumptions before you ask AI for anything. AI works better when the prompt looks more like a specification and less like a vague request.
Generate offline, not online. Use AI to draft blocks, comments, HMI texts or test cases in a development environment. Keep it away from live systems until a human has reviewed the logic properly.
Review like a controls engineer, not like a copy editor. Check sequence logic, retentivity, scan behavior, edge cases, startup conditions, fault states, safe stop behavior, watchdog behavior and operator interactions. If the AI generated motion, safety or comms code, review even harder.
Then test in simulation, emulation or a safe bench setup. Siemens’ own TIA messaging around engineering efficiency also emphasizes testing and simulation functions, which is exactly where AI-generated drafts belong before plant deployment.
Finally, store every change in a controlled workflow with version history. If your project environment can support Git, structured exports and CI-style checks, use them. If not, at minimum keep disciplined offline backups and change logs. AI increases speed, which means it can also increase the speed of bad changes unless governance improves with it.
The skills that will matter most now
A lot of engineers ask whether AI will reduce the need for PLC programmers. That is the wrong question.
The better question is: what kind of PLC programmer will be more valuable in an AI-assisted world?
The answer is an engineer who can do five things well:
understand the process,
define requirements clearly,
spot mistakes quickly,
test methodically,
and secure systems properly.
Deloitte’s 2025 smart manufacturing survey found that adapting workers to the “Factory of the Future” was a top concern for 35% of respondents. ISA’s industrial AI guidance also puts workforce readiness and upskilling near the center of the conversation. That lines up with what many engineers are already feeling: the role is not disappearing, but it is getting broader. The future controls engineer is part programmer, part systems thinker, part validator and part cybersecurity gatekeeper.
Why OT security has to be part of the AI conversation
This part cannot be skipped.
Industrial automation teams are being asked to move faster, connect more assets and adopt smarter tools, while the threat environment stays ugly. IBM X-Force says manufacturing was the most attacked industry in 2025 for the fifth year in a row. Kaspersky’s ICS CERT reported that 21.9% of ICS computers had malicious objects blocked in Q1 2025, with phishing pages, malicious scripts, email threats and internet-delivered threats all still prominent.
That means every AI discussion in OT should include basic questions like:
Where is the data going?
What project information is being exposed?
Can prompts leak customer logic or plant details?
Who reviews generated code?
How are changes approved?
Can the tool be isolated from sensitive production systems?
ISA points directly to ISA/IEC 62443 as an important framework for aligning AI deployment with OT security requirements. That is the right mindset. AI in automation is not just a productivity tool. It is part of the control environment, and it should be treated with the same seriousness as any other engineering system that can influence operations.
What this means for Siemens and TIA Portal users
For Siemens-focused engineers, the direction is very clear. TIA Portal V21 is improving traceability and workflow integration through Git-friendly exports, while Siemens’ industrial-copilot strategy is pushing AI deeper into engineering, operations and maintenance. The message from Siemens is not subtle: future automation work will be more software-defined, more AI-assisted and more connected across the value chain.
That does not mean every Siemens engineer should rush into every new AI feature. It means now is the time to build the habits that make AI useful: cleaner project structures, better naming, clearer specifications, stronger testing, disciplined versioning and tighter cybersecurity boundaries. The engineers who already work that way will get the most benefit first.
Final thought
AI is not magic, and it is not a replacement for engineering judgment. But it is becoming a real force in industrial automation. The winners will not be the people who treat it like a toy, and not the people who dismiss it as hype. They will be the engineers who learn how to use it as a serious tool inside a serious workflow.