AI adoption: do the same words mean the same thing to everyone?
A team can use the same AI tool while assigning different meanings to the information it provides. The management question is precise: how can we ensure that an AI-generated summary preserves the distinctions each professional group needs in order to act?
What the literature establishes
Susan Leigh Star and James Griesemer (1989, Social Studies of Science) describe how actors from different social worlds cooperate through “boundary objects” without sharing all the same knowledge. These objects remain recognisable across groups while accommodating different local uses. This historical study does not address AI, but it encourages us to distinguish between a shared document and an identical interpretation.
When transmission is not enough
Paul Carlile (2002, Organization Science), in a study of new product development, shows that professional knowledge is situated in practices and connected to different interests. Across occupational boundaries, a useful shared artefact must enable people to represent knowledge, discuss differences and transform that knowledge. For AI adoption, the mechanism to examine is therefore the preservation of professional distinctions, not merely the fluency of the text.
The premature conclusion
One might conclude that imposing a glossary before deploying AI is enough. Yet the same term may legitimately reflect different constraints across professional groups: commercial “availability” does not necessarily mean technical availability. The aim is not to eliminate these differences, but to make visible those that change a decision. (our executive and employee training programmes)
What the evidence does not establish
Neither study measures the fidelity of AI-generated summaries or the effectiveness of a glossary in producing them. They provide a framework for analysing cooperation, not proof that a particular protocol will improve a current tool. Nevertheless, a review limited to style and factual corrections risks leaving ambiguities between professional groups unchecked.
A practical check in Sion
As part of SHR’s programme on leading AI adoption within a team, an exercise in Sion could use anonymised or fictional handover documents from hydropower, healthcare, winegrowing and Alpine tourism. Before generating a summary, representatives of two relevant professional groups record the distinctions essential to the decision, such as the difference between advertised capacity and capacity that can actually be deployed. They then check each distinction in the summary and calculate the proportion preserved unambiguously, retaining the source document, the output and the assessment grid. Comparing a generic instruction with one that explicitly states these distinctions, using the same cases and tool, provides a locally verifiable result without establishing its generalisability. To go further: explore the Leading AI Adoption in Your Team training in Sion in the canton of Valais, or browse our executive and employee training programmes in Switzerland.
In pictures: Leading AI Adoption in Your Team in Sion
- leading ai adoption in your team
- Sion
- research
- training sion