AI: does translating an instruction convey its meaning?
In a bilingual organisation, AI can be used to translate a procedure or rephrase a technical instruction for another department. The management question is specific: how can we check that recipients understand the action required, rather than just the words used?
What the literature establishes
Paul R. Carlile (2002, Organization Science) shows, in a study of new product development, that knowledge is embedded in occupational practices. Its circulation is therefore not simply a matter of transferring information: differences in knowledge and dependencies between actors need to be represented and discussed. For an AI-translated instruction, this suggests a risk to investigate, not a demonstrated effect: preserving the terminology does not guarantee that its practical meaning is preserved.
The mechanism: connecting words to a work situation
Beth A. Bechky (2003, Organization Science) describes how occupational communities build shared understanding on a production floor. Concrete objects and situated exchanges help resolve misunderstandings that vocabulary alone cannot overcome. Clear rephrasing therefore does not necessarily replace examining the component, document or problem to which the instruction refers.
The premature conclusion
One might conclude that AI should not be used for bilingual technical communication. These studies do not support that conclusion: they evaluate neither machine translation nor generative models. Instead, they caution against treating a fluent text as evidence that two occupational groups will infer the same course of action. (our executive and employee training programmes)
What the evidence and proofreading cannot guarantee
These qualitative studies illuminate mechanisms of understanding across occupations, without measuring an error rate applicable to translations produced by today's AI. Nor do they isolate the effect of language from that of training, role or experience. Linguistic proofreading remains useful, but it does not directly test a recipient's ability to decide what to do in a given situation.
A practical check in Fribourg
As part of SHR's « Manager à l'ère de l'intelligence artificielle » programme, an exercise could use a fictional instruction for handling a traceability discrepancy in Fribourg's agri-food sector, involving production and quality managers from bilingual industry and, if possible, a partner from the local university community. Working from the same scenario, French- and German-speaking participants read their respective AI-assisted versions, then separately identify the action to take, the person responsible to contact and the condition for resuming operations. The measure is the proportion of responses matching an operational assessment grid validated in advance by the relevant subject-matter leads; each discrepancy is recorded alongside the relevant passage and retested after clarification. This test remains outside production and checks understanding in context, rather than certifying a procedure's safety. To go further: explore the Managing in the Age of Artificial Intelligence training in Fribourg, or browse our executive and employee training programmes in Switzerland.
In pictures: Managing in the Age of Artificial Intelligence in Fribourg
- managing in the age of artificial intelligence
- Fribourg
- research
- training fribourg