AI: does revising a proposal make it harder to abandon?
AI supplies an initial proposal; the team corrects it, expands it and eventually defends it. The precise management question is this: does revision only improve the proposal, or does it also make abandoning it more difficult?
What the literature establishes
Michael I. Norton, Daniel Mochon and Dan Ariely (2012, Journal of Consumer Psychology) show, in assembly and construction experiments, that participants can place greater value on objects they have made themselves. This “IKEA effect” depends in particular on successful completion of the task: effort alone does not always produce this increase in valuation. These findings concern objects rather than proposals written with AI, but they provide a basis for a hypothesis about attachment to outputs we have helped shape.
When past effort enters the decision
Hal R. Arkes and Catherine Blumer (1985, Organizational Behavior and Human Decision Processes) document the tendency to continue an endeavour after investing money, time or effort. The sunk-cost mechanism differs from valuing one's own creation: it concerns the weight given to a past investment in a decision about the future. In AI-assisted work, citing revisions already made as a reason to retain a proposal would be a signal to examine, not evidence of its suitability.
The hasty conclusion
One might conclude that teams should be prevented from revising AI proposals. That would confuse a risk of attachment with the genuine usefulness of revision, which can introduce professional constraints missing from the initial output. The useful distinction is between arguments about the proposal's current quality and arguments about what it has already cost. (our executive and employee training programmes)
What the evidence does not establish
These studies establish neither the existence nor the magnitude of an attachment effect in co-writing with generative AI. A team that retains its revised version may simply have corrected important flaws; an external reviewer may also be unaware of decisive constraints. A preference for one's own version is therefore insufficient to diagnose bias without examining the quality of the options and the information available.
A practical check in Lausanne
Within SHR's “Managing in the Age of Artificial Intelligence” programme, a proposed exercise in Lausanne could use a non-sensitive project brief tailored to a medtech company, a higher education institution, an international sports organisation or a Lake Geneva region scale-up. Before revising with AI, the team sets quality criteria, then compares its final version with an alternative prepared separately from the same requirements; uninvolved colleagues assess both versions without knowing their origin. The measure is the proportion of cases in which the team and those colleagues select different versions, accompanied by their reasons and the recorded revision time. This disagreement does not prove bias, but makes the discussion testable: are people defending a better-satisfied requirement, or mainly the work already invested? To go further: explore the Managing in the Age of Artificial Intelligence training in Lausanne in the canton of Vaud, or browse our executive and employee training programmes in Switzerland.
In pictures: Managing in the Age of Artificial Intelligence in Lausanne
- managing in the age of artificial intelligence
- Lausanne
- research
- training lausanne