Artificial intelligence significantly accelerates standards research – but it does not make the final decision. At CERTAIN, the proven engineering process remains at the center; AI is used to provide targeted support.

Researching applicable standards is one of the most time-consuming steps in the CE process. For the Machinery Directive alone, hundreds of standards are harmonized, their status changes constantly, and selecting the right standards is a key factor in determining whether a machine is considered safe and compliant. It is no wonder that AI is seen as a major lever here. However, a question that often gets lost in the marketing noise is at least as important: what can you rely on – and what can you not?

What AI is actually good at in standards research

As a search assistant, AI plays to its strengths where large volumes of publicly available standard references need to be evaluated and structured quickly:

• Searching through lists of references in seconds based on product features and search terms, instead of reviewing them manually.
• Suggesting relevant A, B, and C-type standards for a specific machine type based on officially published standard titles.
• Flagging published amendments, withdrawn standard references, and their documented successor versions.
• Visualizing cross-references between European legislation and harmonized standard references.
• Drastically reducing the mechanical search effort, thereby freeing up time for technical assessment.

The foundation consists of the public standard references and metadata captured in CERTAIN, as well as the user's input.

This is where AI is a true accelerator. It relieves engineers of tedious search tasks and provides a broad, traceable starting point. The subsequent evaluation is performed by a human: the engineer reviews the suggested standards in full text and decides which requirements apply to the specific case. This human-in-the-loop principle is an integral part of the process — the AI researches, the expert takes responsibility.

Where the engineer remains indispensable

This is exactly where serious application separates itself from over-hyped marketing. Research provides candidates – the decision remains a matter of professional judgment:

• Selecting the appropriate C-type standard is a matter of interpretation, not just a search. Often, there is no perfectly fitting standard, or there may be several competing ones. Which one applies to a specific machine depends on machine limits, intended use, and reasonably foreseeable misuse.

• The presumption of conformity is tied to the official listing status in the Official Journal of the EU and to transition periods. This status changes – classifying it correctly is delicate and directly relevant to safety.

• If a suitable C-type standard is missing, requirements must be derived from A and B-type standards as well as the state of the art. No tool can assume this responsibility.

• AI can provide plausible-sounding but incorrect or outdated standard information. Without professional verification, this false sense of precision is dangerous.

• The manufacturer, not the software, bears the responsibility for conformity. "The AI suggested it" is not a valid justification, neither before an authority nor in the event of a liability claim.

Especially during the transition from the Machinery Directive to the Machinery Regulation, the landscape of standards is in flux. This requires human interpretation, not just a search result.

Human-in-the-loop: How CERTAIN uses AI

CERTAIN uses AI intentionally as an assistant, not an autopilot. The principle is simple: the AI provides well-founded suggestions, but the final decision always rests with the human.

• Suggestions, not decisions: The engineer confirms, rejects, or supplements—with full traceability at all times.

• Method over tools: The iterative risk assessment process according to EN ISO 12100 provides the framework. AI accelerates individual steps but does not replace the methodology.

• Traceability: Every decision is documented via version control—a significant advantage for audits and compliance checks.

This process has evolved from years of CE practice. It remains at the core; the AI is a tool within this proven process, not a replacement for it. That is part of our DNA.

Why being honest about limitations is a mark of quality

Those who are open about the limitations of their tools are more trustworthy than those who promise omnipotence. In a liability-sensitive field like CE marking, this is not a flaw, but a mark of quality. CERTAIN uses AI where it demonstrably helps—and remains transparent about where engineering judgment is required.

Final thoughts

AI is changing standards research for the better—when used correctly. As an accelerator, it relieves the burden of routine tasks and creates space for what really matters. Interpretation, responsibility, and the final judgment remain with the human. It is precisely this interplay between a proven process and targeted AI support that makes the difference between an impressive demo and reliable conformity.

Would you like to see how CERTAIN combines AI with engineering methodology? Discover the platform at certain-cloud.com.