AI on the job: an available answer is no substitute for identifiable expertise
An AI assistant can become a team’s first point of contact before the team has clarified the role of its experts. The management question is specific: how can knowledge become more accessible without losing the ability to identify the right person when an answer requires context-sensitive judgement?
What the literature establishes
Kyle Lewis (2003, Journal of Applied Psychology) developed and validated a measure of transactive memory systems in work teams. This concept describes a collective organisation of knowledge: members hold different areas of expertise, recognise their credibility and coordinate their use. Collective work therefore depends not only on available information, but also on knowing how knowledge is distributed among people.
The mechanism: also learning who knows what
Diane Wei Liang, Richard Moreland and Linda Argote (1995, Personality and Social Psychology Bulletin) compared group training with individual training before a group assembly task. In this experiment, groups whose members had trained together performed better, with findings supporting the mediating role of transactive memory. Learning together can thus build an awareness of people’s expertise that simply providing instructions does not guarantee.
The premature conclusion
One might infer that an AI assistant necessarily weakens this mutual knowledge. These studies do not examine generative AI and do not support that conclusion. Instead, they suggest a distinction to test: a tool can provide an answer, but it can also point to a source and to someone capable of assessing its local relevance. (our executive and employee training programmes)
The limits of the evidence and current practice
An assembly experiment and the validation of a team-level measure do not establish the effects of an AI assistant in a bank or a fashion business. Moreover, having an expert directory proves neither that employees know how to use it nor that the people listed are available. Measuring only how quickly an answer can be accessed would therefore overlook the organisation’s ability to handle exceptions.
A practical check in Lugano
For an SHR “AI on the job” workshop in Lugano, we propose using fictional cases written in Italian: a banking documentation exception, a material specification in fashion, a delivery term in trading or delegated decision-making in an Italian-speaking family SME. For each case requiring internal expertise, ask the participant to find a relevant source and a competent contact using approved tools, then have the process owner confirm that assignment. The measure is the proportion of cases for which both the source and the contact are confirmed, compared before and after explicit references are added to the materials used with AI. This check does not measure the tool’s entire value; it tests whether access to answers comes with effective access to expertise. To go further: explore the AI on the Job training in Lugano, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Lugano
- ai on the job
- Lugano
- research
- training lugano