Managing early-career staff has always involved teaching judgement, and now the spread of AI tools makes that responsibility more important and more complex. AI is changing how we complete tasks and how we organise work. We have heard from managers and leaders through AI CoLab  it is starting to shift how tasks and tasking are explained, which is also shifting how work is reviewed, and how (or what) feedback is important to give.

There are concerns about whether AI adoption is changing early-career development. Jobs and Skills Australia’s Our Gen AI Transition Case Studies(Opens in a new tab/window) found people earlier in their career may have fewer opportunities to learn through the formative tasks that once built judgement over time. For managers, some practical questions include:

  • How do I help staff use AI while still building their judgement?
  • How do I notice which learning moments are being changed by AI?
  • How do I keep development visible when the work itself is changing? 

What correction used to teach 

In the past, a lot of early-career learning happened through correction. Usually, this meant a draft coming back loaded with comments, deletions and tracked changes. It was often laborious and confronting, but over many iterations people gradually stumbled towards something resembling skill or at least built up the muscle needed to keep their supervisor happy. 

A newer staff member drafted something, a supervisor reviewed it, and the document moved back and forth through comments, corrections and conversations. It could be slow and frustrating, but it also made the thinking visible. 

That process was not perfect, and it was rarely fast. But it forced people to spend time talking through the work together, because deliberate communication and critical reflection were often the only way to reach a good final product. 
 

What is AI shifting about this experiential learning?  

AI use changes the balance because a polished draft can now appear very quickly. Managers can often produce or rewrite the work themselves faster than they can explain the task, teach the reasoning, review the draft and work through revisions with someone else. 

That creates a real efficiency gain, yet it also removes many of the moments where people were forced to slow down and learn how to think through the work. 

An April post by Anthea Roberts ‘Learning Agency: Two Processes, Not Just One(Opens in a new tab/window)’  reflected on the harder work of helping people learn through the process, rather than letting AI quietly compress that process into a polished final output. In a team setting, that means making prompts, choices, errors, corrections and judgement visible. 

When AI helps move work from a rough idea to polished draft quickly, it can hide the steps where learning often happens:

  • framing the task
  • weighing options
  • spotting risks, and  
  • deciding when the output is not good enough. 

Track changes are less useful when a whole draft can be regenerated in seconds. Even generous and thoughtful managers who spend time reviewing drafts line by line still need new ways to teach effective AI use. 

How can supervisors adapt?  

The old game has flipped. Learning used to happen partly by osmosis because it was built into the process of getting to the final product. Now there may be faster ways to reach the same objective, but we still need to build and, in many cases, discover effective ways to teach the judgement needed to do that work well. 

That is where management judgement still matters most, because the real task is often helping staff understand what the work is asking beneath the surface: 

  • who it is for,  
  • what decision it supports,  
  • what the history and context are,
  • what risks sit around it, and  
  • what would make the answer useful in context.  

AI may help produce the words faster, but human management and attention are what’s needed to teach people how to read and interpret the task properly before they start prompting. 

Here are three approaches for teams and managers to start experimenting with as we learn how to use these tools while still teaching judgement, context and good decision-making.
 

Provide feedback on the process and approach, in addition to end products 

A useful shift is reviewing how someone used AI, instead of only reviewing the final output. In addition to asking whether the draft is good, the conversation becomes:  

  • What prompts were used?
  • What context was included or left out?
  • Where was the tool helpful?
  • Where did it miss the point?
  • Where did a person need to step in? 

The change in focus is from simply reviewing the output to making the manager’s tacit judgement more visible. A lot of the real learning sits in the small decisions people make along the way, especially when an AI-generated answer looks polished but quietly misses the point. 

This also reflects broader observations from MIT Sloan and Stack Overflow that AI is reshaping workflows(Opens in a new tab/window) and making shared knowledge, context and workplace learning more important. 
 

Build the first prompt together 

Shift the focus to spend time at the start of the task, before anyone produces a polished draft. Managers and staff can work through the first prompt together by discussing the audience, the risks, the context, what a useful answer needs to do, and what could make the work fail. 

This front-loads some of the thinking that used to happen later through track changes, rewrites and back-and-forth edits. Much of management judgement is implicit by default. A manager may instinctively know that a ministerial audience needs less detail, that a stakeholder issue is sensitive, or that a technically correct answer will still fail because it misses the institutional context. The task shift is to make that judgement explicit earlier, so human input can be applied where it adds the most value: framing, context, sensitivity and decision quality. 

Using AI responsibly forces more assumptions into the open. The quality of the output depends heavily on whether the context was made explicit in the first place, so more of the reasoning and judgement can be surfaced earlier through discussion, instead of being slowly absorbed through multiple rounds of editing. There is still room for ambiguity, and work will always be reviewed. 
 

Encourage leadership 

Remember that everyone is learning!  
 

Microsoft Australia’s Ctrl + Career report on Gen Z and AI adoption(Opens in a new tab/window) found early-career staff often have more time to experiment, fresh eyes on existing processes, and a willingness to try new tools and workflows before everyone else feels ready. Teams can harness that energy to test new ways of drafting, organising information, reducing repetitive work, and improving everyday processes together. 
 

Where do I start? 

An easy way to start is to drop this post into a tool like Copilot and ask something like: 

Based on this article, suggest five ways public sector managers and early-career employees can work together to build judgement, risk awareness and professional capability while using AI. Include practical discussion prompts we can use as a team. 

You can read more from these guiding sources