AI assistants: how can teams preserve diversity of ideas?
No-code assistants can quickly suggest ideas for an offering, a communication campaign or a service improvement. The managerial question is specific: should employees consult the assistant before developing their own ideas?
What the literature establishes
Anil R. Doshi and Oliver P. Hauser (2024, Science Advances) experimentally examined access to AI-generated ideas in a short-story writing task. Access improved creativity ratings for individual outputs, particularly among participants who were initially less creative, but the AI-assisted stories were more similar to one another. Improvement in individual outputs can therefore coexist with a reduction in diversity across the collection.
The mechanism of fixation on an example
David G. Jansson and Steven M. Smith (1991, Design Studies) showed in design tasks that exposure to an example could lead participants to reproduce its features, including problematic ones. An example does not merely provide a resource: it can also narrow the space being explored. This mechanism offers a possible lens for understanding an assistant’s suggestions, without directly establishing its effect in a team using AI.
The premature conclusion
It would be tempting to conclude that AI should be excluded from creative activities. Instead, the findings distinguish two objectives: improving a proposal and maintaining a range of different proposals. Collecting individual ideas before displaying the assistant’s suggestions is an organisational hypothesis to test, not a method validated by these studies. (our executive and employee training programmes)
What the evidence cannot settle
Short stories and design exercises reproduce neither the constraints of a healthcare service nor the trade-offs facing a hydropower operator. These studies also do not establish the long-term effect of a no-code assistant shared by a team. Finally, counting different formulations is insufficient: several texts may describe the same operational option.
A practical check in Sion
As part of SHR’s programme « Créer ses assistants et agents IA sans coder », a workshop in Sion could compare two sequences using fictional cases: consulting the assistant before individual idea generation, or only afterwards. Cases could cover information about hydropower works, administrative reception in healthcare, a vineyard visit and an alpine tourism offering, without sensitive data or safety decisions. With equal time and equal numbers of proposals, evaluators unaware of the sequence used would group duplicates and count genuinely distinct options that meet predefined feasibility criteria. Repeating the comparison with the sequences reversed between groups can inform a local organisational choice without claiming to establish a general rule. To go further: explore the Building AI Assistants and Agents Without Coding training in Sion in the canton of Valais, or browse our executive and employee training programmes in Switzerland.
In pictures: Building AI Assistants and Agents Without Coding in Sion
- building ai assistants and agents without coding
- Sion
- research
- training sion