No-code AI assistants: who maintains them when work changes?
The programme “Creating your own AI assistants and agents without coding” makes it possible to build tools tailored to an activity. One managerial question remains: who should update the assistant when a procedure, a role or a workflow changes?
What the literature establishes
Wanda J. Orlikowski (1992, Organization Science) proposes a theoretical framework, illustrated by a field study, in which technology is both a product of human action and an element that structures that action. Its effects therefore arise not only from its technical characteristics, but also from its uses and institutional context. For an AI assistant, this framework invites scrutiny of both its initial configuration and the practices that keep it aligned with work.
A written procedure is not the whole routine
Martha S. Feldman and Brian T. Pentland (2003, Administrative Science Quarterly) distinguish the general representation of a routine from its concrete performances by people in particular situations. Their theoretical analysis explains how this relationship can generate stability as well as change. Applied to AI assistants, it suggests that a tool configured around a procedure may remain unchanged while the actual handling of requests evolves.
The premature conclusion
One might conclude that the assistant should immediately be adapted to every observed variation. That would conflate a local exception, a questionable workaround and an organisational change that has actually been approved. The managerial task is instead to decide which changes should be incorporated into the tool, by whom and from what date. (our executive and employee training programmes)
What these studies do not demonstrate
These publications address neither generative models nor no-code platforms; they do not measure the effect of a maintenance policy on their performance. They provide an analytical framework, not an optimal review schedule. Checking that an assistant works technically is nevertheless insufficient to establish whether it still follows the current allocation of tasks and recipients.
A practical check in Sion
In a SHR programme workshop in Sion, assemble a set of fictional requests: a hydropower maintenance report, a healthcare administrative enquiry, a vineyard tasting booking and a change to an alpine tourism service. For each case, record the expected recipient role and workflow, then introduce an approved organisational change and assign the update to a named owner. Run the affected cases again and calculate the proportion in which the assistant follows the new workflow: the number of compliant cases divided by the number of affected cases tested, retaining dated outputs. This measure does not certify the assistant's overall quality; it checks whether a specific change in work has actually been incorporated. 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