Generative AI can turn a natural-language software requirement into usable code in seconds. It is not a controls engineer; it is a remarkably fast tool for turning project requirements into a structured program or function.
What If PLC Programming Started With a Sentence?
Article from | Unitronics
“Start the pump when the tank level falls below 35%. Stop it at 80%. If the motor trips, stop the sequence, latch an alarm, and require a manual reset. Prevent an immediate restart.”
It sounds like a line from a machine specification or a controls meeting. It can also be the beginning of a PLC program.
Generative AI can turn a natural-language software requirement into usable code in seconds. It is not a controls engineer; it is a remarkably fast tool for turning project requirements into a structured program or function.
That distinction matters in industrial automation. PLC code controls physical equipment. It must behave deterministically, enforce interlocks, handle faults, and respond safely to failed sensors or changing operating conditions. No engineer would hand those decisions to a chatbot.
Where does AI fit in?
Much of PLC development is not high-level engineering judgment. It is translation and repetition: converting machine specifications into sequences, I/O lists into tags, operating modes into states, and fault requirements into alarms. Engineers also spend hours on declarations, repetitive functions, comments, and documentation.
This is the grunt work AI can dramatically reduce. Every hour saved can shorten development cycles, lower engineering costs, and free skilled people for work that genuinely requires their expertise.
AI is already generating measurable value in manufacturing. Deloitte’s AI in Manufacturing 2026 reports that 84% of more than 140 manufacturers surveyed are realizing measurable value from AI; 40% use generative or agentic AI.
That value can now extend to control programming.
Integrating AI output into PLC Control
AI cannot generate Ladder Logic--but it can generate functions that can be called from a Ladder application. Large language models work naturally with text-based programming languages, and can generate Structured Text (ST), an IEC 61131-3 textual language suited to calculations, loops, arrays, data handling, state machines, reusable functions, and more. ST is therefore a natural bridge between a machine description and executable control logic.
Engineers can keep Ladder where they prefer it, for example for discrete control, interlocks, and online troubleshooting--and can save time by integrating AI-generated ST functions into the same project.
How to generate usable PLC functions with AI
Your AI service may know Structured Text syntax, but it will not automatically know the context—your PLC platform’s supported data types, project structure, naming rules, functions, or import requirements. It also does not know which tags, I/O, structures, and functions already exist in your project.
Without that context, generated code can create almost as much cleanup as it saves.
Unitronics addresses this problem through the AI Wizard in UniLogic, the programming environment for UniStream controllers. The workflow connects the engineer’s requirement with the actual project context before code is generated:
- The engineer describes the required application or function in ordinary language, or attaches a specification.
- Relevant project elements, including I/O, tags, structures, and existing functions, are selected as context.
- UniLogic creates an application-specific prompt and platform instructions for the engineer’s chosen AI model, such as Claude or ChatGPT.
- The engineer reviews the generated package and imports it into UniLogic. For a complete application, this package can include structures, tags, functions, enumerations, initial values, and a main program.
The generated output also marks places where the specification or code still needs human attention. “Assume” identifies assumptions made because information was missing. “To Do” identifies decisions or work that remain open. “Fix Me” flags unsupported constructs or issues requiring correction. Instead of hiding uncertainty behind polished-looking code, the workflow makes it visible.
What changes for the controls engineer
The immediate benefit is not autonomous PLC engineering.
It is less time spent creating the first version of routine logic, declarations, comments, and documentation. The same wizard-based workflow can help modify or explain existing Structured Text, review a specification for missing conditions, flag potential logic issues for engineering review, and adapt text-based code to UniLogic conventions.
A less experienced engineer can begin with guided questions and see how machine requirements map into Structured Text. A senior engineer can spend less time typing repetitive logic and more time checking architecture, abnormal states, recovery behavior, and commissioning results. But clean-looking generated code can still miss a failed sensor, an operator intervention, a mode transition, or a restart condition. AI provides a starting point, not borrowed experience.
The engineer still owns the machine
PLC code controls physical equipment. Whether AI is used in the design cycle or not, the final responsibility lies with the engineer who must verify:
- Safety architecture, risk assessment, interlocks, and fail-safe behavior
- Electrical design, I/O selection, sequence completeness, and recovery from faults
- Simulation, compilation, bench testing, FAT, SAT, change control, and final approval
- Protection of project data, credentials, intellectual property, and network access when external AI services are used
Natural language does not make the PLC probabilistic. The deployed control still executes deterministic logic with explicit states, interlocks, and tested behavior. AI belongs earlier in the workflow, where requirements are organized and the first code draft is prepared for human judgment.
A new starting point for PLC programming
So, can PLC programming start with a sentence? Yes—but the engineer must still define what the machine should do, what it must never do, and how it should respond when conditions are not normal.
UniLogic AI does not remove engineering from the process. It removes the blank screen. By bringing the requirement, actual project context, and platform rules together, it delivers a reviewable first draft sooner—and makes unresolved assumptions visible from the start.
Built directly into UniLogic and provided at no additional cost, the workflow puts AI-assisted development within reach of machine builders and system integrators of every size.
Less time translating intent into code. More time proving that the machine is safe, reliable, and ready to run. The future of PLC programming does not begin with less engineering. It begins with a better first draft.
The content & opinions in this article are the author’s and do not necessarily represent the views of ManufacturingTomorrow
Featured Product
