AI adoption: is working language distorting the diagnosis in Berne?

In Berne, an AI adoption initiative may bring together colleagues who do not all work in their strongest language. The management question is specific: how can difficulty using AI be distinguished from difficulty with the language in which its use is required?

What the literature establishes

Boaz Keysar, Sayuri L. Hayakawa and Sun Gyu An (2012, Psychological Science) found in their experiments that certain decision biases were reduced when participants reasoned in a foreign language rather than their native language. This does not mean that working in a foreign language generally improves decisions. More modestly, it establishes that the language in which information is presented is not necessarily a neutral vehicle for reasoning.

A language barrier can become a judgement about people

Helene Tenzer, Markus Pudelko and Anne-Wil Harzing (2014, Journal of International Business Studies) conducted a qualitative study of trust formation in multinational teams. Their analysis describes how language barriers shape the perceptions and attributions through which colleagues develop trust. For AI adoption, this raises a hypothesis worth examining: difficulty formulating or discussing a prompt could be mistaken for poor command of the tool.

The hasty conclusion

One might conclude that translating training materials or imposing a common language is sufficient. These studies validate neither solution for AI. A common language can facilitate communication while requiring unequal effort; translated materials can support understanding without making group discussion equally accessible. (our executive and employee training programmes)

What the evidence does not establish

Neither study examines generative AI adoption within a team in Berne. Decision-making experiments and a qualitative study of trust cannot estimate the effect of multilingual support on workplace use. Moreover, directly comparing two prompt languages could confound an employee’s language proficiency with the model’s linguistic performance.

A practical check in Berne

As part of SHR’s « Conduire l'adoption de l'IA dans son équipe » programme, a Berne-based team in federal administration, public health, telecommunications or precision manufacturing can test two support arrangements using comparable fictional or non-sensitive tasks. Each participant completes one task with the usual instructions and another with instructions and support in their strongest working language, reversing the order for some participants while keeping the model’s prompt and output language unchanged. The main measure is the proportion of tasks completed without additional help and meeting a predefined quality rubric, assessed without knowledge of the support arrangement used. A difference would not establish a general language effect, but it would provide verifiable evidence before particular employees are labelled reluctant AI users. To go further: explore the Leading AI Adoption in Your Team training in Bern in the canton of Bern, or browse our executive and employee training programmes in Switzerland.

In pictures: Leading AI Adoption in Your Team in Bern

• leading ai adoption in your team
• Bern
• research
• training bern