Managing with AI: what do we sacrifice for immediate performance?
A team can use AI to improve current results while devoting less effort to understanding what might transform them. The management question is specific: how can we assess an AI application without systematically favouring immediate returns over learning?
What the literature establishes
James March (1991, Organization Science) distinguishes the exploitation of existing knowledge from the exploration of new possibilities. His model shows how the more immediate and predictable returns from exploitation can lead an organisation to give it excessive weight. Applied to AI management, this framework invites us to distinguish improvements to a familiar process from investment in knowledge that remains uncertain.
The mechanism: learning within too short a horizon
Daniel Levinthal and James March (1993, Strategic Management Journal) analyse the myopias of learning, including the tendency to favour consequences that are close in time and within the organisation. A practice can therefore be reinforced because its local benefits are visible, while its more distant costs remain outside the scope of assessment. By analogy, a dashboard focused on time saved through AI risks leaving abandoned trials and uninvestigated questions invisible.
The premature conclusion
It would be excessive to conclude that AI necessarily impoverishes learning or that exploration should always take priority. Exploiting a proven method remains essential, particularly when repeatability and compliance determine quality. The qualification concerns the criteria used to make trade-offs: rewarding only immediate results does not reveal a team's future ability to solve a new problem. (our executive and employee training programmes)
What these studies do not allow us to claim
These two articles offer theoretical frameworks, including a simulation model; they do not evaluate today's generative AI tools. They therefore provide neither a quantified effect of AI on learning nor an optimal proportion of time to devote to exploration. Counting completed trials is also insufficient unless their content, results and conditions for reuse are documented.
A practical check in Neuchâtel
In a Neuchâtel team working in microtechnology, watchmaking or precision microelectronics, an AI pilot could focus on preparing root-cause analyses of non-conformities without changing validation procedures. Before the pilot, and then over a comparable period during it, record planned investigations outside production and calculate the proportion that yield a documented result, including a negative one, noting the reasons for abandoning any investigation. Reading this indicator alongside the time spent on analyses and the complexity of the cases could reveal a possible shift in effort without automatically attributing it to AI. This offers a concrete exercise for SHR's “Manager à l’ère de l’intelligence artificielle” programme: making visible what performance assessment usually leaves out. To go further: explore the Managing in the Age of Artificial Intelligence training in 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