AI on the job: does greater use mean a better fit for the task?
After AI training, a manager may be tempted to track logins or prompts to assess the team’s progress. The more precise management question is how to check whether the tool suits the tasks assigned, rather than simply establish that people are using it.
What the literature establishes
Fred Davis (1989, MIS Quarterly) shows that perceived usefulness and perceived ease of use are associated with the acceptance of information technology. These perceptions help explain why someone adopts a tool; they are not an independent measure of its contribution to the work. A satisfaction questionnaire therefore provides information about acceptance, without being sufficient to establish suitability for professional tasks.
The fit between task and technology
Dale Goodhue and Ronald Thompson (1995, MIS Quarterly) propose and test a model in which performance impacts depend, among other factors, on utilisation and the fit between task and technology. This fit concerns the capacity of available functions to support work requirements. Applied to AI, the framework directs attention to a specific activity rather than assigning a general value to the tool.
The premature conclusion
One might conclude that increased use after training proves its success. Yet it could also reflect a managerial instruction, easy access or the appeal of novelty: these are possible explanations, not findings established for generative AI by these two articles. Conversely, infrequent use may be appropriate when it is reserved for tasks that suit the tool. (our executive and employee training programmes)
What common indicators cannot establish
These studies concern information systems that predate generative AI and draw partly on self-reported measures; they do not demonstrate the effects of a current deployment in precision manufacturing. Prompt counts likewise cannot distinguish useful assistance from a succession of unsuccessful attempts. Managers therefore need to supplement adoption indicators with observation of the requirements actually met.
A practical check in Neuchâtel
Within SHR’s “IA on the job” programme, an exercise could focus on preparing shift handover instructions at a Neuchâtel company in microtechnology, watchmaking or precision microelectronics, using fictitious or explicitly authorised data. Over a predefined period, compare instructions prepared with and without AI for comparable situations, then have them checked against the same task-specific criteria: equipment concerned, operation status, any anomaly and required action. The measure would be the proportion of instructions usable without a request for clarification, reported alongside raw counts and preparation and correction time. This local check would not prove a general effect of AI, but would support a discussion of its fit for a task rather than its usage rate alone. To go further: explore the AI on the Job training in Neuchâtel, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Neuchâtel
- ai on the job
- Neuchâtel
- research
- training neuchâtel