We Didn’t Just Automate the Jobs. We Removed the Years.
For most of my career, the entry-level job was not really a job.
It was a laboratory. It was the place where someone who knew nothing about professional life learned how to think: how to sit in a meeting room and read what wasn’t being said, how to handle a difficult brief from someone who expected better, how to recover from a mistake without the floor falling out beneath them. The entry-level job taught the hidden curriculum. The one that doesn’t appear in any syllabus but shapes everything you do afterward.
I’ve been watching this pattern for a long time. When I wrote about the future of apprenticeships in a post-automation world in 2024, the argument was that the entry-level role wasn’t a cost to be optimised but a pipeline to be protected. The piece explored what happens when organisations strip away structured exposure at the beginning of a career, not just to the person who loses the position but to the whole development chain above them.
What I didn’t fully name then, and what we are all beginning to understand now, is that AI wasn’t going to take the senior jobs first.
It was going to take the first years.
And the first years are where everything gets built.
What Entry-Level Work Was Actually Doing
Before we can name what’s being lost, we need to name what it was.
Entry-level work was never about efficiency. An inexperienced analyst completing a task took three times longer and made twice as many errors as a senior one. Every organisation knew this. They hired graduates anyway, because the inefficiency was the investment.
The young accountant who reconciled the wrong ledger and had to explain why. That was a lesson in precision, in accountability, in reading what a partner’s silence actually meant about professional standards.
The junior copywriter who submitted a campaign brief and watched a creative director tear through it line by line. That was an education in craft, in what language actually does when it lands on a real audience.
The graduate intake who attended a board meeting and wrote up the minutes. That was an apprenticeship in organisational politics, in how power moves through rooms, in what gets said versus what gets decided.
None of this appeared in a job description. It was the invisible curriculum that turned graduates into professionals.
It took time. It took roughly two to five years of dense, structured exposure before the investment began to pay back significantly. The organisation knew it was paying for that developmental window. The person didn’t always know they were inside it. Both benefited enormously.
That accumulation (the professional instinct, the judgment, the capacity to read a situation quickly and act well) doesn’t come from a training module. It comes from doing the work badly and being corrected, repeatedly, at close range with someone whose standards you’re learning to internalise.
That is what’s at stake. That is what the AI entry-level jobs transition is dismantling, quietly, one efficiency decision at a time.
The Archaeology of a Learning Pipeline
There’s a pattern here that goes back further than most people realise.
The formal apprenticeship, where a young person was bound to a master craftsman for years of structured exposure, wasn’t a quaint pre-industrial arrangement. It was a technology for transferring tacit knowledge across generations. The master didn’t just teach technique. They transmitted judgment. The apprentice didn’t just learn tasks. They absorbed a whole way of seeing and deciding.
The industrial revolution compressed and cheapened this model. When work was broken into repeatable units, the need for the full apprenticeship declined. You could train someone to do one step in a production line in days, not years.
Law, accounting, medicine, architecture, consulting: the professions largely preserved the apprenticeship model through a different mechanism, the junior position. The first few years of a professional career were structurally designed to work like an apprenticeship, even when nobody called them that. Volume work under supervision. Close exposure to senior judgment. Explicit feedback cycles. Time.
What AI is doing to the professional entry-level function is structurally similar to what the industrial revolution did to the craft apprenticeship. It is inserting a highly capable non-human agent at exactly the point in the pipeline where human learning was supposed to happen.
The craft apprenticeship survived the industrial revolution in specialised form. The professional learning pipeline may not survive this transition in its current form without conscious redesign.
The Signals That Were Already Moving
In July 2025, I wrote a piece called Majoring in Obsolescence about education systems still training students for a world that no longer exists. The line that cut through most sharply was this: the traditional university major was a frozen idea in a streaming world.
The same logic applies to the entry-level role as it has been structured for the last fifty years.
The signal has been visible across several sectors for some time now. Graduate hiring has softened in legal, financial services, consulting, and media: the sectors with the highest concentration of cognitive entry-level work. Research from the World Economic Forum and leading labour economists has consistently pointed toward the same pattern: AI’s near-term displacement is concentrated not at the top of the skill distribution, where human judgment and relationship management remain dominant, and not at the very bottom, where physical presence and manual dexterity are still required, but in the middle-entry zone: the cognitive tasks that require competence but not yet mastery.
That is precisely the zone where young professionals were supposed to be doing their developmental time.
These are Immediate Futures. Not speculation. Not projections for 2035. Things already moving through the system, already reshaping who gets hired, when, and for what.
The pace at which AI tools are absorbing first-draft writing, document review, data processing, initial research synthesis, and client communication templates isn’t a coming disruption. It’s a present one. And the organisations responding to it primarily with efficiency metrics are missing the second signal underneath.
What This Means for a Specific Person
Here is what this means for a 24-year-old with a law degree entering a mid-tier firm in 2026.
A significant portion of the work they expected to do in their first two years has been absorbed by AI tools and efficiency redesign. What remains is either higher-order work they aren’t yet equipped to do without significant supervision, or operational tasks that don’t build transferable professional instinct at the pace the old model did.
They are not reading and annotating three thousand contracts. They are reviewing AI outputs and flagging exceptions.
This is not a useless activity. It builds a different kind of skill: the ability to evaluate AI-generated work critically, to spot the error that looks plausible, to apply a professional standard to material that wasn’t produced by a human mind.
But it is not the same developmental path as doing the original work.
The professional instinct (the one that comes from doing enough volume of lower-stakes real work to have a gut sense of what something unusual looks like) doesn’t load the same way. The capacity to draft a persuasive document from first principles doesn’t develop at the same rate when you spend your formative years editing AI outputs rather than producing originals.
This person will arrive at year four or five of their career technically employed and credentialed. They will have performed well against the updated benchmarks. And there will be something in their professional foundation that is thinner than it would have been if they’d come up five years earlier. Something they may not be able to name, something their managers will only notice when the work is genuinely difficult and the scaffolding has to come from inside rather than from the tool.
That thinness won’t show in normal conditions.
It will show in the hard ones.
The PTFA of Talent Systems
This is where PTFA (Past Trauma, Future Anxiety) enters the frame.
Organisations have been here before. The redundancies of the 2008 financial crisis stripped out layers of middle management that were never fully rebuilt. The teams that remained were smaller, flatter, and running harder. The institutional memory and mentorship capacity that used to sit in those layers went with the headcount.
When the next generation of junior professionals arrived into those organisations, the scaffolding they expected wasn’t there. Senior people were too stretched to be genuine mentors. The feedback loops were thinner. The development happened but more slowly and with more gaps.
Organisations that went through that carry the scar. Some of them shaped better mentorship structures afterward. Many of them never quite closed the capability gap in the cohorts that came up in the thin years.
What’s happening now is structurally similar but arrives through a different mechanism. It isn’t redundancy. It’s substitution. The role is still there. The salary is still there. But the developmental core of the role has been replaced by something that can do the tasks more efficiently but can’t do the thing that made the tasks worth doing.
The future anxiety runs in the other direction too. Leaders can see what AI is making possible. The efficiency argument is compelling and in many cases correct. The pressure to demonstrate AI-enabled productivity gains is real. And sitting between the scar of the last restructure and the promise of the current efficiency play, many organisations are making the same decision without naming it as a decision about talent development at all.
That is the trap.
Ripple Effects: Who Trains the Leaders of 2035?
This is the second-order consequence. And it is the one that matters most strategically.
In seven to ten years, the senior professionals of 2035 will have been shaped by this compressed developmental window. The partners, the directors, the heads of practice (the people making calls that cost organisations significant money) will have come up through a system that removed or substantially reduced the dense early-career exposure that built professional judgment in previous generations.
The skills gap won’t appear at recruitment. It will appear when the pressure is on and the judgment call is genuinely difficult and the person who’s supposed to make it has the title, the credentials, and the confidence, but not quite the depth.
Consider the third-order consequence: who will mentor the generation that comes after them?
Professional mentorship depends on the mentor having lived through the experiences they’re helping someone else navigate. If the mentor’s own early career was shaped by reviewing AI outputs rather than producing original work under senior scrutiny, they will pass on a different set of intuitions. Not worse ones necessarily. Different ones. Shaped by a different developmental reality.
The pipeline doesn’t just thin at the entry point. It thins at every point downstream that depends on what was learned at the entry point.
A decision that appears as an efficiency win in 2024 shows up as a capability problem in 2031.
This is how Ripple Effects work. Not as single consequences but as cascading shifts through time, through culture, through the human systems that carry organisational capability across generations.
The Tension Inside Organisations Right Now
Most organisations I work with are holding two contradictory beliefs simultaneously, and they know it.
The first: AI can handle the entry-level cognitive work, and letting it do so makes the organisation demonstrably more efficient and competitive.
The second: We need a talent pipeline. We need to develop people. We cannot indefinitely hire senior capability from outside and expect it to carry the culture and institutional knowledge that comes from having been inside.
Both beliefs are true. They are also in direct tension with each other.
The organisations that will do this well are the ones that name the tension explicitly and design for it, rather than letting the efficiency argument win by default simply because it is louder.
I explored some of the early signals of this tension in a piece examining the AI versus Gen Z hiring debate: the question of what early-career professionals bring that AI can’t replicate, and what organisations are actually buying when they hire a junior professional in an AI-enabled environment.
The answer isn’t the tasks. The tasks are increasingly acquirable from the tool.
What’s being bought is the person’s capacity to develop. Their ability to accumulate judgment over time. Their future senior self.
If the organisation isn’t designing for that future self, if the role isn’t genuinely building the person toward something, then the hire is a different kind of investment than it used to be. And the organisation owes itself an honest conversation about what it is actually trying to produce.
What Immediate Futures Demands
This is not an argument against AI in the entry-level function. I have never been in that camp and I’m not starting now.
The question isn’t whether to use AI for entry-level cognitive tasks. Many of those tasks are better done by AI: faster, more consistent, less susceptible to the errors that come from fatigue and distraction. The question is what you are doing in parallel to replace the developmental exposure those tasks used to provide.
Some models I’ve seen working:
Deliberate exposure design. Organisations that have explicitly redesigned the junior pathway to include structured volume work on lower-stakes real problems: not AI outputs to review, but original work to produce. The AI handles efficiency at the system level; the junior professional gets a carved-out territory where they build genuine instinct through doing rather than through evaluating.
Human mentorship as the core architecture. In the apprenticeship piece from 2024, I argued for what I called the blended learning model, where digital tools and human mentors work in genuine partnership, with the mentor’s role explicitly redesigned around judgment transfer, not task supervision. This becomes more important, not less, as AI handles more of the task layer. The mentor is the developmental architecture. The AI is the productivity infrastructure. These are different things and they need to be designed separately.
Hiring for development capacity rather than task readiness. If the entry-level role has genuinely changed, the person who thrives in it has different attributes than the one who thrived in it in 2019. Not necessarily more technical. More self-directed, more comfortable with ambiguity, more capable of learning from evaluating AI work critically. The hiring profile needs to reflect what the role actually builds toward, not what it used to require on day one.
Explicit institutional knowledge preservation. The organisations most at risk are the ones that have allowed AI efficiency gains to quietly reduce the senior-to-junior ratio without redesigning how the institutional knowledge that used to live in those ratios now gets transmitted. That knowledge doesn’t disappear because the headcount structure changed. It goes somewhere, often into the heads of people too stretched to transmit it deliberately.
None of this is impossible. All of it requires the deliberate decision that the organisation’s talent pipeline is worth designing for, not just hoping the market will supply the senior capability when it’s needed.
The Preparation Window
The problem with thinning a developmental pipeline is that the consequences arrive on a delay.
The organisation that reduces its graduate intake significantly in 2025 and 2026 will not feel the consequence in 2027. The consequence arrives in 2031 and 2032, when the reduced cohort would have been at the stage of meaningful senior contribution. By that point, the decision that created the gap is three or four budget cycles in the past. The causal link is hard to name. The organisation scrambles to hire from outside, which is expensive and imperfect and doesn’t rebuild the institutional depth.
The preparation window is the time between now and when the gap becomes undeniable.
That window is not unlimited. And it is closing faster than most talent and workforce planning functions currently have in their models.
The organisations that treat this as an Immediate Futures question, something already arriving and already requiring a response, will have options that the organisations waiting for clearer evidence won’t.
We didn’t just automate the jobs.
We removed the years that the jobs were supposed to provide.
And the work of designing what replaces those years is exactly the work that most organisations haven’t yet named as urgent.
It is urgent.
Choose Forward.
Morris Misel is a foresight strategist and keynote speaker based in Melbourne, Australia. With 30+ years of experience working with leaders, boards, associations, and organisations across Australia and internationally, he works with people to prepare for uncertainty, interpret signals, and make better strategic choices. Learn more: morrisfuturist.com
Frequently Asked Questions
What does AI taking entry-level jobs mean for professional development?
When AI absorbs the cognitive tasks that entry-level professionals used to do (document drafting, initial research synthesis, data processing, first-pass analysis) it removes the structured exposure that built professional judgment. The risk is not just fewer jobs but thinner development: people arrive at senior positions having accumulated different and in some ways shallower professional instincts than previous cohorts who did the volume work by hand.
Which industries are most affected by AI displacing entry-level learning pipelines?
The sectors most exposed are those with the highest concentration of cognitive entry-level tasks: legal, financial services, management consulting, media and communications, and accounting. These are fields where the first two to five years of a career involved high volumes of cognitive work under senior supervision: exactly the zone AI tools are most capable of absorbing.
How can organisations protect talent development pipelines in an AI environment?
Organisations can redesign junior pathways to include deliberate exposure to original problem-solving rather than purely AI output review. Human mentorship needs to be explicitly positioned as the developmental architecture, not a side function. Hiring criteria should shift toward development capacity (self-direction, learning agility, critical evaluation) rather than task readiness alone. The institutional knowledge that used to live in senior-to-junior ratios needs a new transmission mechanism.
What is the long-term organisational risk of removing entry-level developmental exposure?
The senior professionals of 2030-2035 will have been shaped by a compressed developmental window. The capability gap won’t show at recruitment. It will surface when genuinely difficult judgment calls are required and the depth built through early-career volume work isn’t there. This is a Ripple Effect: an efficiency decision made in 2025 produces a capability problem in 2031 that is difficult to trace back to its origin.
What does Morris Misel’s Immediate Futures framework mean in this context?
Immediate Futures refers to what is already arriving and needs attention now, not speculation about distant possibilities but signals already moving through the system. The decline in AI entry-level jobs and the learning pipeline disruption is an Immediate Futures condition, not a future one. Graduate hiring has already softened across exposed sectors. The developmental redesign needs to happen now, during the preparation window, before the gap becomes undeniable.
