Getty Images
AI ate the first rung of the technology career ladder
As AI automates routine tasks in tech careers, it could deprive junior professionals of learning opportunities. And it may create a workforce that lacks experience and expertise.
For years, the technology industry has reassured itself that AI will not replace people. It will merely do the routine work, leaving experienced humans to supervise it, correct its mistakes and exercise the judgment machines lack.
This sounds reassuring, particularly if you are one of the experienced humans. But it leaves a fairly substantial question unanswered: Where will the next generation of experienced humans come from?
Technology careers have traditionally operated as informal apprenticeships. Developers became senior developers by writing code -- some of it inelegant, some of it wrong and much of it patiently corrected by somebody more experienced. Support technicians learned by fixing straightforward problems before being trusted with complex incidents. Project managers learned their craft by running smaller projects whilst discovering that a plan and reality are often only nodding acquaintances.
AI now consumes many of those formative tasks.
A junior developer can ask an assistant to generate code, a service desk tool can diagnose and resolve common faults, and an AI project assistant can construct the plan, summarise meetings, update risks and chase actions. Each use case has an attractive financial justification because work is completed faster, fewer people are required and expensive senior staff can supposedly concentrate on higher-value activities.
The problem is that we may be automating not only the work, but also the learning.
The disappearing apprenticeship
The commercial logic is difficult to resist. If one experienced developer with AI can produce work that previously required three junior developers, a spreadsheet somewhere will inevitably recommend employing one experienced developer. It is much harder to place a value on the two senior developers the juniors might have become in five years' time.
Businesses have encountered this problem before. Organizations that cut graduate and apprenticeship programs during downturns often find themselves short of managers and specialists several years later. AI heightens the temptation because eliminated work can really disappear, while the resulting expertise gap remains conveniently outside the current budgeting cycle.
Checking is an expert skill
The usual answer is that humans will remain in the loop. However, this phrase conceals another problem because properly checking an AI-generated answer often requires much of the expertise needed to produce it.
A developer who has not developed much code may not recognize a subtle security weakness in a plausible-looking output. A support analyst who has mainly accepted automated recommendations may struggle when several systems fail in an unfamiliar combination. A project manager who has never negotiated scope, challenged an unrealistic deadline or recovered a failing project may be able to maintain an excellent AI-generated plan without understanding why delivery is collapsing around it.
Novices can therefore produce expert-looking work without acquiring expert reasoning. Worse, polished AI output can create an illusion of competence for both employees and managers. There are endless stories of complete faith in generative AI overtaking the critical-thinking effort required to challenge the machine.
Thinking has not disappeared, but it has shifted toward verification and integration. These are valuable skills, but only when the reviewer has enough independent knowledge to challenge the machine, as well.
If we change nothing, we risk creating an oddly inverted workforce in which increasingly inexperienced people nominally supervise increasingly capable technology.
Redesign the ladder
This does not mean CIOs should preserve obsolete work as a sort of corporate heritage attraction, nor should junior developers be required to hand-code everything because their predecessors suffered through it. AI can accelerate learning as well as undermine it.
Studies of customer support agents have found that an AI assistant increased productivity by 14% overall, and by 34% among novice and lower-skilled employees. The tool appeared to transfer some of the behaviors of high performers to less-experienced colleagues. Used as a tutor, critic and collaborator, AI could help apprentices move up the learning curve faster, but this requires deliberate job design.
We should consider models where juniors might form an initial diagnosis before consulting AI, explain why generated code works, manually test it against unusual conditions, shadow complex incidents or run controlled projects and then defend their decisions to experienced colleagues.
Progression should measure independent judgment, not simply the speed and polish of AI-assisted output. Mentoring may need to increase precisely when the immediate need for junior production decreases.
The future expert may write less code, personally resolve fewer routine tickets and manually maintain fewer project plans. Their expertise will lie increasingly in defining problems, understanding systems, challenging assumptions, handling exceptions and accepting accountability. None of those capabilities appear automatically when somebody is given an AI license and promoted after five years.
The uncomfortable choice for technology leaders is therefore not between adopting AI and protecting entry-level jobs; it is between using AI to redesign the apprenticeship and using it to abolish the apprenticeship.
Eliminating junior roles may deliver irresistible savings today, but if every organization consumes experienced talent without replenishing it, the industry will eventually discover that it has removed the first rung of the ladder while still expecting people to appear at the top.
