A potted plant remains tied to a stake with a blue tie; higher up, another blue ribbon hangs from the stake alone.

The path to making decisions with autonomy

Onboarding progresses when someone can take responsibility for decisions with judgment, resources and authority. Seeking advice can be part of that autonomy, not its negation.

Alejandro Mayorga7 min read

Imagine an order coordinator who has just joined a distributor. She knows the catalog, has completed orientation and uses the system correctly. A client asks for an earlier delivery. There are units in the warehouse, and the screen allows the date to be changed, but some of those units are reserved for another order. Before responding, the coordinator asks for help. That moment admits two readings: she still depends on someone to do her job, or she recognized a decision for which she does not yet have sufficient judgment. The distinction matters because each reading leads to a different way of supporting her.

If the person helping simply resolves the case, the order may turn out well and the next exception will need the same intervention again. If they show her where to check reservations, they will have passed on part of the missing knowledge. But she still needs to understand which commitments can be changed, what alternatives to offer and who is authorized to accept that a client will wait. Learning to operate the screen and learning to answer for that decision follow related, not identical, paths.

This is a hypothetical situation, but it helps clarify what we expect of onboarding. Providing information allows someone to find their bearings; it does not demonstrate that they can already interpret differences between cases or exercise the role’s authority. Nor is it enough for their first deliverables to be correct: those may have been possible because someone else anticipated every difficulty. To speak of autonomy, we need to know which part of the work the person can lead, with what resources and within what limits.

Research on organizational socialization helps prevent that path from being reduced to content. The review and meta-analysis by Bauer and colleagues, published in 2025 (opens in a new tab), distinguishes role clarity, task mastery, social acceptance and perceived fit as dimensions of adjustment. It finds associations between support from organizational members and those indicators. It establishes neither a universal sequence nor a deadline for becoming autonomous. It does allow us to recognize that understanding the work and finding a place from which to participate in it are problems an information-based orientation does not exhaust.

Someone may arrive with extensive experience and need help understanding what an exception means here. They are not starting their profession over. They are learning to situate their judgment within commitments, relationships and rules they do not yet know. Interpreting every request for advice as a technical deficiency can lead us to teach what they already know while leaving implicit precisely what they need to learn.

The judgment a demonstration does not reveal

Let us return to the order. An experienced person may check reservations before changing the date, but that check may go unnoticed by an observer: the expert knows which screen to open and which data to disregard. If the demonstration concentrates on the sequence of actions, the coordinator will see a smooth solution without understanding which signal would have required stopping it. Expert execution can hide part of the knowledge we want to pass on.

The cognitive apprenticeship model of Collins, Brown and Holum (opens in a new tab), presented in 1991, proposes making those thought processes visible, supporting practice and gradually withdrawing support that is no longer needed. It was developed in the teaching of reading, writing and mathematics, not as a validation of corporate onboarding programs. The relevant relationship here is pedagogical: observing the result is not the same as observing how it was reached. In our example, it would be necessary to explain why other orders are checked and what difference turns a date change into a reallocation of commitments.

For that learning to happen, error cannot be the only way to discover a consequence. When a decision may harm third parties or is difficult to reverse, a reconstructed case or a proposal reviewed before execution allows judgment to be developed without transferring the cost of practice to them. Nor is this about demanding certainty before every intervention. It is about creating room to decide where risk can be controlled, while retaining review that still serves a necessary function.

There is also a condition no amount of practice replaces: access to information and authority. If reservations are not accessible to the role, asking for better judgment will not make them appear. If the coordinator must request permission for every modification, even after demonstrating that she can resolve it, a rule is preserving the dependence. Withdrawing support without changing those conditions leaves the person exposed; maintaining every approval while demanding autonomy sets an expectation they have no means to meet.

The questions that remain necessary

Seeking advice also changes function along the way. At the beginning, it may mean asking someone to say what to do. Later, it may serve to test an interpretation or bring a consequence beyond the role’s scope to the appropriate authority. The coordinator may be prepared to resolve ordinary modifications and, precisely because of that, recognize when an order requires a different decision. Autonomy in one domain does not turn limits of authority into personal deficiencies.

Not every question demonstrates good judgment: some may show that learning remains incomplete or that the same difficulty has not been worked through. But counting questions without examining what is being asked confuses those cases with a relevant consultation. It can also make another problem invisible. In Amy Edmondson’s study of 51 teams in a manufacturing company (opens in a new tab), psychological safety is associated with learning behaviors, including asking for help and discussing mistakes. It is not causal evidence about onboarding. It does, however, alert us to a condition of our interpretation: what someone dares to ask also depends on how they expect us to receive the question.

A person with a clipboard observes a loading dock with boxes on pallets, a forklift and a truck.

That is why checking that fewer questions now arrive would not be enough. In the coordinator’s case, we would want to reconstruct different decisions: a modification she resolved, a proposal she corrected on receiving new information and a request she escalated while explaining which commitment was at stake. Reviewing the result matters too. A convincing justification does not compensate for having ignored a reservation, just as being right by chance does not demonstrate understanding of the case. These observations help delimit what can become her responsibility without turning a favorable sample into permission to resolve any exception.

Whoever organizes onboarding needs to devote time to that path and update what remains under supervision. A schedule helps the review happen; it does not itself produce the capability it expects to find. It may be reasonable to close general orientation while specific support continues, or to expand authority in one domain before another. What ceases to be reasonable is maintaining indefinitely a provisional condition nobody reviews, or declaring someone autonomous solely because the allotted time has ended.

Onboarding should enable someone to take responsibility for a domain of decisions, including those that require asking for help. Those of us providing support are responsible for making judgment visible and granting the authority the person can already handle. If all we achieve is that they stop asking us questions, we do not know whether we developed autonomy or taught them to disguise dependence.