AI on the job: does personalised feedback help people improve?
A manager can ask AI to comment on a presentation, a document or a simulated interview. The managerial question is not whether the feedback sounds personalised, but whether it helps the employee perform the next task better.
What the literature establishes
Avraham N. Kluger and Angelo DeNisi (1996, Psychological Bulletin) show in a meta-analysis that feedback interventions do not consistently improve performance and can also impair it. Their theory proposes, in particular, that effectiveness depends on where feedback directs attention: towards the task or towards concerns about the self. Detailed commentary is therefore not, by itself, a tool for improvement.
The mechanism: directing attention towards action
John Hattie and Helen Timperley (2007, Review of Educational Research) distinguish feedback about the task, processes, self-regulation and the person. Their review frames useful feedback in terms of the relationship between a goal, the current state of the work and the actions needed to close the gap. For AI use, this suggests requesting a diagnosis tied to an explicit criterion rather than a general portrait of an employee’s qualities.
The premature conclusion
One might conclude that asking AI for specific, actionable comments is sufficient. But a concrete recommendation can still be wrong, peripheral or impossible for the recipient to implement. Personalised wording guarantees neither a relevant diagnosis nor a feasible course of action. (our executive and employee training programmes)
What the evidence cannot promise
These publications do not examine today’s generative assistants, and the review by Hattie and Timperley primarily concerns learning in educational settings. They offer a framework for designing feedback, not proof that AI-generated feedback is effective in the workplace. Measuring only recipient satisfaction or the quality of the writing would leave improvement itself outside the assessment.
A practical check in Lausanne
As part of SHR’s “IA on the job” programme, an exercise in Lausanne could focus on a recurring task: explaining a technical result in medtech, reviewing a project at a higher education institution, preparing a dossier for an international sports federation or presenting an offer at a Lake Geneva region scale-up. Using fictional or authorised materials, compare usual feedback with feedback prepared using AI and validated by the manager, then have participants complete a new, comparable task without assistance. Define an observable criterion beforehand, such as a verifiable justification for every recommendation, and have the work assessed by someone who does not know which type of feedback was received. The relevant measure would be the proportion of outputs meeting this criterion in each condition, rather than ratings of the feedback itself. To go further: explore the AI on the Job training in Lausanne, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Lausanne
- ai on the job
- Lausanne
- research
- training lausanne