AI: should teams have usage targets?
A dashboard can show that employees use AI without establishing that their work is improving. The managerial question is specific: should teams be asked to increase their frequency of use, or their ability to identify appropriate uses?
What the literature establishes
Steven Kerr (1975, Academy of Management Journal) uses organisational examples to analyse the gap between rewarded behaviours and officially desired outcomes. His article does not address AI, but identifies a problem directly relevant to its deployment: rewarding a visible activity does not guarantee the intended result. A target for logins or prompts may therefore reward tool use regardless of its usefulness.
The mechanism: shifting effort towards what can be measured
Bengt Holmström and Paul Milgrom (1991, Journal of Law, Economics, & Organization) model incentives when work involves multiple tasks whose contributions are not equally measurable. Their analysis shows why strengthening incentives on an easily observed dimension can divert effort from other important dimensions. Applied to AI, this mechanism raises the question of whether a usage target leaves sufficient room for choosing not to use the tool.
The premature conclusion
It would be excessive to conclude that every adoption indicator is harmful. User numbers can help identify access difficulties or support needs without becoming an individual performance criterion. The useful distinction is between a diagnostic indicator and a target tied to rewards, comparisons or managerial pressure. (our executive and employee training programmes)
What these studies cannot establish
Neither article measures the effects of generative AI usage quotas: one offers an illustrated organisational analysis, the other a theoretical model. They therefore establish neither an optimal frequency nor a quantified loss of quality. In practice, comparisons of usage volumes across functions are also problematic, because relevant opportunities to use AI vary by activity.
A practical check in Neuchâtel
At a Neuchâtel company specialising in microtechnology, watchmaking or precision microelectronics, an exercise within the SHR programme “Manager à l’ère de l’intelligence artificielle” could involve reviewing adoption targets before communicating them to teams. During a pilot period defined in advance, record the use or reasoned non-use of AI for each eligible case, then record whether the deliverable passes its first review against the usual quality criteria. Calculating the proportion of deliverables accepted without rework, separating comparable tasks with and without AI, makes it possible to compare usage frequency with a business outcome. This comparison does not establish causality, but it does allow managers to check whether increased use actually coincides with an improvement in the selected quality indicator. To go further: explore the Managing in the Age of Artificial Intelligence training in Neuchâtel in the canton of Neuchâtel, or browse our executive and employee training programmes in Switzerland.
In pictures: Managing in the Age of Artificial Intelligence in Neuchâtel
- managing in the age of artificial intelligence
- Neuchâtel
- research
- training neuchâtel