The Apprenticeship
Gap.
When AI performs the junior work, who becomes senior?
Every profession has a room nobody photographs.
It is where the junior lawyer reads a bad contract and learns why one clause is dangerous. Where the AML analyst works through a false positive until they can tell a pattern from a coincidence. Where the reporter makes twenty calls for one paragraph. Where the developer breaks a small thing, traces the cause and learns what the system was doing before it failed.
The work is slow, repetitive and poorly paid. It is also how a person becomes the senior person everyone later wants to hire.
AI is exceptionally good at this layer. It can produce the first brief, the first research pass, the first compliance summary, the first code review, the first customer reply. That is why companies are adopting it. The immediate saving is visible. The missing apprenticeship is not.
The Work Nobody Loves Is the Work That Teaches
There is a habit in management language: divide work into “high-value” and “low-value,” then automate the second category. It sounds efficient because it describes the task only from the buyer’s side.
From the worker’s side, much of that supposedly low-value work is not an output. It is a training environment. A junior accountant does not reconcile transactions because reconciliation is the final destination. They do it until irregularity becomes visible. A paralegal does not read bundles because reading is glamorous. They do it until evidence has weight, sequence and consequence.
Remove every first pass and you may improve this quarter’s productivity. You may also remove the path by which judgment enters the organisation.
Fewer routine hours, faster drafts, more output per experienced employee.
The hundreds of small decisions from which instinct, standards and responsibility are built.
The Paradox Is Not That AI Helps Novices
The best evidence is more interesting than the panic. In a large study of customer-support workers, access to a generative AI assistant raised productivity on average, with the largest gains for novice and lower-skilled workers. The tool captured patterns from stronger workers and made them available in the moment.
That is not a case against AI. It is evidence of its real promise: a junior worker can receive a useful suggestion at the point of need instead of waiting for an expert to become available.
The question arrives one step later. Does the worker learn why the suggestion is good? Can they handle the exceptional case the assistant has not seen? Does the organisation preserve time for explanation, review and independent attempts — or does it treat the higher output as proof that less training is required?
The Missing Middle
Every organisation thinks it needs seniors. It actually needs a functioning middle.
The middle is formed by people who have done enough ordinary work to recognise the extraordinary. They are the ones who notice that an answer is polished but wrong, that a customer is not asking the question on the ticket, that a risk score is being gamed, that a system has passed every automated check and is still unsafe.
AI can make a small senior team unusually productive. It cannot, by itself, create the next layer of people capable of auditing that team. If entry-level work becomes a narrow exercise in prompting, accepting and escalating, the pipeline eventually delivers people with outputs but without a base.
A Small Country Cannot Afford a Hollow Pipeline
This matters everywhere. It matters differently in a small country.
Large economies can sometimes hide a weak talent pipeline behind scale. They can import specialists, absorb a failed training cohort, or let one industry compensate for another. Malta does not have that luxury in every field. Its financial services, gaming, compliance, legal, technology and media sectors depend on people who understand both the rulebook and the reality in front of them.
That is why access to AI education matters — including Malta’s public effort to make structured AI learning available to eligible residents. Access is a good beginning. But access is not an apprenticeship. A completed course does not replace the junior who must make a difficult call, defend it, be corrected and return the next day sharper.
The opportunity for Malta is not simply to distribute tools. It is to become unusually good at building human oversight around them: practical training, accountable deployment and careers in which AI raises the floor without removing the ladder.
The Wrong Response Is to Protect Busywork
No serious answer is to preserve inefficient work just because it once trained people. Nobody should defend a career built on copying data between systems or spending nights formatting a document that software can format in seconds.
The answer is to identify what the old work was secretly teaching, then design a better path. A junior lawyer can use AI to find cases, but must still construct the argument and explain why the authority applies. A compliance analyst can use a model to organise evidence, but must write the rationale for a decision and face review. A developer can use an assistant to propose code, but must trace the system, test the edge cases and own the incident when it breaks.
That is harder than simply cutting junior roles. It is also what leadership looks like when a business intends to exist beyond the current tool cycle.
The Contract Between Generations
An apprenticeship is a quiet contract. The junior gives effort, humility and time. The senior gives access to judgment. The institution absorbs some inefficiency because it is investing in a future it cannot yet measure.
AI changes the economics of that contract. It gives the institution a tempting alternative: keep the senior, buy the model, remove the waiting room. The decision may be rational at the level of one department. Across an industry, it is a bet that expertise can continue to appear without the conditions that create it.
That bet has never been safe.
The question is not whether a junior can use AI. They should. The question is whether, after using it, they can still explain the work, challenge the answer and perform the essential part when the system is unavailable or wrong.