AI adoption: should teams protect time for exploration?

When a team finds a satisfactory use for AI, its manager has good reasons to consolidate it. One question remains: how much room should be left to explore other uses when immediate results favour those the team already masters?

What the literature establishes

James G. 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 overemphasise it, to the detriment of long-term adaptation. Applied to AI, this distinction cautions against confusing the refinement of an established use with an expansion of the team’s possibilities.

The mechanism: learning can narrow the search

Daniel A. Levinthal and James G. March (1993, Strategic Management Journal) analyse the myopias of learning, including the tendency to neglect distant horizons and favour what experience makes locally visible. A team can therefore become more proficient in its current practices without learning more about alternatives it does not try. For managers, the risk is not merely a lack of experience with AI, but allowing the first successful uses to define the entire scope of subsequent experimentation.

The premature conclusion

One might conclude that teams should multiply experiments and distrust routines. These studies do not justify that conclusion: exploitation also helps stabilise quality and capitalise on previous learning. The issue is a trade-off between two necessary activities, not a systematic preference for novelty. (our executive and employee training programmes)

What the evidence cannot determine

These articles offer a model and a theoretical analysis of organisational learning; they do not measure contemporary generative AI adoption. They establish neither an optimal share of time for exploration nor an ideal number of uses to test. A dashboard restricted to the volume of AI-assisted output is nevertheless insufficient to document this trade-off: it leaves possibilities that were considered and rejected unmeasured.

A practical check in Sion

As part of the SHR — Swiss Human Resources programme « Conduire l’adoption de l’IA dans son équipe », a team in Sion could keep a one-month log distinguishing improvements to an existing use from investigations of a new one. In hydropower, healthcare, viticulture or Alpine tourism, exploration would focus on non-critical document-related tasks, using authorised data and human validation. The proposed measure would be the share of testing time devoted to each category, with a documented result and a reasoned decision to continue or stop for each new avenue. This record would not prove the value of exploration, but it would establish whether exploration has a real place or remains merely an intention. 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