Early in my career, I had a job called “Office Angel”.
I still have the description. Travel and logistics for more than 400 presentations a year. Briefing meetings, contracts, calendars. There was also a requirement to be a “beacon of love” for speakers travelling around the world. An ambitious addition to an administrative role, but there it was.
The work involved making sure people got where they needed to be, with what they needed, and felt cared for along the way. Looking back, I can see how much of my professional life began there.
My mentor, Jules, taught me far more than the mechanics of the job. I absorbed something of her mindset, attitude and instinct. Her approach became a reference point for my own, often without me noticing.
It took years to recognise how much had stayed with me. The questions I asked. The things I paid attention to. The decisions I made when there was no obvious answer. I followed a different career path and made those lessons my own, but her influence travelled with me.
The job description listed what I would do. It could never have captured who I would become.
Junior work has long performed this double duty: producing something useful while developing the person doing it. You research a problem, write a draft, miss something important. Someone more experienced explains what you overlooked. The next time, you notice a little more.
As AI takes on more of those tasks, we need to ask what happens to the learning within them. When an organisation changes the work, how deliberately does it consider the person still learning to do it?
Getting a start
The first signs of change may appear in who gets through the door.
In August 2026, Stanford researchers reported that employment among Americans aged 22–25 in AI-exposed occupations was around 19 per cent below where it would have been had it kept pace with less-exposed peers. The gap came mainly from reduced hiring rather than increased departures. The authors describe these findings as early indicators, not proof that AI caused the gap, and found no widespread economy-wide displacement.
Australia offers a different picture. Indeed Hiring Lab reported in April 2026 that graduate job advertisements had stabilised above pre-pandemic levels. The timing of earlier declines suggested AI adoption was not the primary driver. These are different measures in different labour markets, but they are a useful reminder that there is no single, inevitable trajectory.
Nor can hiring figures tell us everything about development. They do not show whether a young person has a good mentor, gets meaningful feedback or leaves a working day understanding something they did not understand that morning.
A junior position can provide all of those things. It puts you in the room. You hear an experienced colleague ask an unexpected question. You discover why a client’s brief and what they actually need can be different things. You make a decision, see what happens and begin to develop judgement of your own.
Getting a start matters. But even where junior jobs remain, we need to ask what opportunities to learn remain within them.
What the work was teaching
Imagine a graduate and an experienced colleague reading the same AI-generated recommendation. Both can follow its argument. The experienced colleague notices an assumption that does not hold for this particular client. The graduate has never encountered that situation before.
Recognising a problem often depends on having something to compare it with: an earlier mistake, a difficult conversation, an approach that failed. The experienced colleague has a foundation to draw on. The graduate is still building theirs.
The Thomson Reuters Institute has described a professional judgement gap: the risk that automating formative tasks removes experiences through which emerging professionals learn to make sound decisions. It is a concern about what could be lost as education and work change, rather than proof that a generation’s judgement is already declining.
The question for employers is practical. When a task disappears, what did someone learn from it, and where will they learn that now?
Sometimes the answer will be that the task taught very little. Repetitive administration, without explanation or feedback, was never automatically a good education. My experience with Jules does not mean everyone had a Jules. Plenty of people have spent years doing junior work without receiving much help to move beyond it.
That is a reason to build something better. AI could be part of how we do it.
A better way to begin
Research on customer-support workers, published in The Quarterly Journal of Economics, found that an AI assistant increased productivity by about 15 per cent on average, with larger gains for novice and lower-skilled workers. The researchers also found evidence of learning. One workplace study cannot settle the question for every profession, but it shows why we should take AI’s potential to help beginners seriously.
A young employee could use AI to explore an unfamiliar concept, rehearse a conversation or compare approaches before discussing them with a colleague. Work previously beyond their reach could become their next useful challenge.
The design matters. Completing a task with assistance and understanding it well enough to take responsibility for it are different achievements. A good workplace should help people develop both.
If AI drafts the report, a graduate could investigate its assumptions, check its evidence and explain their recommendation to a mentor. If research takes less time, the time gained could support a discussion about why credible sources disagree. If routine administration shrinks, beginners could gain earlier access to clients and decisions, with support to prepare and reflect.
“Check the AI’s answer” is a thin instruction for someone still learning what a good answer requires. They need opportunities to hear another person’s reasoning, test their own and understand why a decision changed.
Looking back at my Office Angel years, the tasks gave me responsibility. Working alongside Jules helped me grow into it. Technology could make that combination available to more people, provided we make learning part of the plan.
And the ability to benefit from those opportunities begins developing long before the first day at work.
Before the first payslip
A child explaining how they reached an answer is practising reasoning. Students working through a disagreement are learning to listen, question and collaborate. A university student revising a recommendation after feedback is learning that changing your mind can improve your work.
These experiences build on one another. Later, the same capabilities help a new employee ask for clarification, recognise the limits of their knowledge or speak up when something seems wrong.
Adults contribute through their own behaviour, too. A teacher admits they had not considered another possibility. A parent acknowledges a mistake. A mentor explains why they changed a decision. Each makes a way of thinking visible to someone developing their own.
Schools can strengthen these foundations alongside academic knowledge. Universities can create opportunities to apply them through supervised projects and professional experience. Employers provide access to the clients, constraints and consequences that formal education cannot fully reproduce.
Those responsibilities are connected, but they are not interchangeable. A school cannot decide whether a company hires beginners. A university cannot give an employee’s manager time to teach. Employers control those choices.
A young person’s development crosses the boundary between education and employment. Our responsibility needs to cross it too.
An ambition worth having
This is the broader proposition behind Stanford’s Learning Society initiative: learning should be supported throughout life, with investment in people shared across the institutions and organisations that benefit from their development.
That requires a longer view than the next qualification or recruitment round. Mentoring takes time. Beginners need support. The full value of helping someone develop may emerge years later, sometimes in another organisation. Yet if every employer wants experience and fewer are willing to provide it, the problem eventually belongs to everyone.
At a Stanford convening on early-career pathways, Professor Mitchell Stevens suggested that AI could represent a Sputnik moment. The Soviet Union’s launch of the first satellite prompted sustained US investment in education, science and human capability. Stevens argued for a comparable commitment to people in response to AI.
The ambition is useful because it asks more of us than preserving yesterday’s jobs. It asks what opportunities we could build now.
An employer introducing AI could put a development plan alongside its efficiency targets: which learning experiences are changing, what will replace them and who will take responsibility? Schools, universities and employers could build projects with real clients, paid placements with proper supervision and mentoring relationships that last beyond a single careers event. Industry bodies and governments could help fund pathways that individual organisations struggle to sustain alone.
We should judge those efforts by whether young people gain meaningful opportunities and become more capable through them. Faster output is valuable. So is a beginner who can explain a decision, recognise a mistake and take on a harder problem next time.
Jules’s guidance became part of how I work. I could not have predicted its value from a list of responsibilities, and I did not fully recognise it while it was happening. Its influence emerged over years.
That is the timescale we need to hold alongside the immediate gains from AI. The young person learning to make a difficult decision today may be the experienced colleague someone else depends on in ten years.
The future of education and the future of work are the same conversation. We are investing in what machines might become. We can bring the same ambition to young people.
Somewhere, someone is writing a job description. It will list the tasks, the skills and the experience required.
It is worth asking what experience the job will give in return.