AI adoption: does the first number determine the deadline?
When a team uses AI to prepare a schedule, it often receives an estimate before examining the constraints of the work. The management question is specific: does the agreed timeline rest on the case information or on the first number displayed?
What the literature establishes
Amos Tversky and Daniel Kahneman (1974, Science) describe anchoring: an initial value can influence a numerical estimate even when it provides no relevant information. Judgement therefore does not arise solely from the facts of the problem. For a team receiving a numerical AI proposal, this is a risk to investigate, not evidence that every estimate will be biased.
Why revision may not go far enough
Nicholas Epley and Thomas Gilovich (2006, Psychological Science) examine adjustment from anchors that people generate themselves. Their findings indicate that adjustment may stop once a value seems plausible, before moving sufficiently far from the starting point. This mechanism highlights possible limitations of an instruction such as “revise the proposed timeline”, without establishing that AI-generated numbers operate in exactly this way.
The premature conclusion
One might conclude that AI should never be allowed to suggest timelines. That would confuse useful information with an unsupported anchor: an estimate grounded in comparable cases can improve the work. The issue is whether the team examines that grounding or simply negotiates around the number displayed. (our executive and employee training programmes)
What the evidence cannot establish
These studies address neither generative AI nor collective planning in Fribourg businesses. They cannot predict the magnitude of any anchoring effect when employees have operational data, documented constraints and opportunities for discussion. Nor is it enough to observe that a human estimate remains close to the proposed timeline: both may be justified by the same information.
A practical check in Fribourg
Within the SHR programme « Conduire l'adoption de l'IA dans son équipe », an exercise could use a single fictional planning case adapted to food processing, bilingual manufacturing or services within Fribourg’s university hub. Randomly assign participants to two identical versions of the case, differing only in an initial estimate attributed to AI, shorter in one and longer in the other, while balancing working languages. Measure the difference between the median timelines ultimately chosen, then disclose the manipulation during the debrief: a shift in the direction of the initial values would be consistent with anchoring, without proving it on its own in a small group. This check assesses sensitivity to the first number, not the accuracy of the schedule, which would require an independent benchmark. To go further: explore the Leading AI Adoption in Your Team training in Fribourg in the canton of Fribourg, or browse our executive and employee training programmes in Switzerland.
In pictures: Leading AI Adoption in Your Team in Fribourg
- leading ai adoption in your team
- Fribourg
- research
- training fribourg