Context card
AI vs. classical automation
How does AI differ from classical automation?
The short answer
Classical automation executes rules that people have defined in advance — a PLC program, a workflow, an RPA bot — and behaves the same way every time. AI systems derive their behavior from data or models and produce outputs that are probable rather than fixed. That is why AI output in operations is usually bounded by deterministic control, permissions and human approval.
For: Technical writers and engineers new to industrial AI
Key points
- Rule-based automation is deterministic: the same input gives the same result.
- Machine learning learns patterns from data; its outputs carry uncertainty.
- In plants, the PLC and the safety functions stay in charge; AI output becomes a recommendation or a bounded command.
- AI has been part of operations for years; agents that call tools are the newer step.
The context
Rules versus learned behavior
Automation in the classical sense means predefined logic: a programmable logic controller runs its program cyclically, robotic process automation replays defined steps in software. The behavior can be specified, tested and repeated.
Artificial intelligence, and machine learning in particular, derives behavior from data. The result is useful where rules are hard to write — detecting anomalies, reading text, planning — but it is a probability, not a guarantee.
How the two work together
In industrial settings AI output rarely acts directly. It becomes a recommendation, a message or a defined call that deterministic control checks against limits, permissions and safe states; human in the loop steps approve consequential changes.
What it means for documentation
State which function is rule-based and which is learned, where AI output enters the control flow, which limits apply and who approves. Naming a technique does not show that a system meets any requirement.
Questions readers ask next
- Is RPA a form of AI?
- Not by itself. RPA replays defined steps; it becomes AI-assisted only when a learned component, such as document recognition, is added.
- Can AI replace a PLC?
- Not in the control and safety functions. AI can suggest setpoints or detect patterns, while the PLC and safety functions keep executing deterministically.
Review log and changes
Every context card is checked against its sources before it is published, and again whenever it changes; the date under the byline is the last review. Corrections (something was wrong) and additions (something was missing) are logged below with date and time (Berlin time). Typos, formatting and link fixes are not listed.
Reviewed
No corrections or additions since publication.