AI on the job: should AI consultations be grouped together?
In financial services, insurance, tech and the European headquarters of international groups in Zurich, consulting AI can become an extra step between two stages of handling a case. The management question is specific: should these consultations be grouped into dedicated periods rather than interspersed throughout the work?
What the literature establishes
Sophie Leroy (2009, Organizational Behavior and Human Decision Processes) shows that switching from one task to another can leave an “attention residue”: some attention remains attached to the previous task. Her experiments indicate, in particular, that leaving a task unfinished can impair performance on the next one. Applied to AI, this finding suggests a hypothesis to test, not an established effect: making repeated requests about different cases could make it harder to resume work.
Switching tasks requires a change of mental setup
Joshua S. Rubinstein, David E. Meyer and Jeffrey E. Evans (2001, Journal of Experimental Psychology: Human Perception and Performance) identify time costs associated with task switching in experimental settings. Their findings support an executive control mechanism involving goal shifting and activation of the relevant rules. Moving from risk analysis to drafting a message with AI may therefore involve more than changing windows, although these experiments do not quantify the workplace cost.
The premature conclusion
One might conclude that AI should be restricted to fixed time slots. Yet consulting a tool to resolve an issue in the current case does not necessarily constitute a task switch in the sense used by these experiments. The useful distinction is between a consultation that advances the same objective and one that opens up another piece of work; a uniform scheduling rule risks conflating the two. (our executive and employee training programmes)
What the evidence cannot decide
These articles concern neither generative AI nor organisations in Zurich. They cannot establish an optimal frequency of consultation, and the number of switches between applications does not directly measure attention residue. The common practice of counting requests or measuring response speed therefore overlooks the organisation of the work in which they occur.
A practical check in Zurich
As part of “IA on the job”, a Zurich team could run a two-week test on comparable internal cases suitable for processing with its authorised tool, alternating half-days of grouped consultations with half-days of usual use. A log containing no confidential content would record active processing time, switches to another case and returns for correction after review for each case. The comparison would examine median processing time and the proportion of cases requiring correction, comparing similar case types. This local test would not establish a cognitive mechanism, but it would help determine whether grouping consultations is worth retaining rather than prescribing it as a matter of principle. To go further: explore the AI on the Job training in Zurich, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Zurich
- ai on the job
- Zurich
- research
- training zurich