The dull work built your senior people. Now AI does it first.

The traditional way to build a senior professional is to hand a junior one dull, repetitive work and wait. Do it badly, then better, then without thinking – that is where judgement comes from. The tedium was never the point; the point was what accumulated while you slogged through it.

So what happens when AI does the dull, repetitive work first?

It is the question sitting underneath the Asia Tech Alliance’s (ATA) new paper, Making AI Work: A Practical Agenda for APAC Workers and Firms – and the one its co-author, Dr Gary Ang, is most willing to admit nobody has answered yet. “There were no clear answers from the interviews,” he tells HRM Asia, “which is expected, given that this is a discussion that has only come to the fore in the last one to two years.”

What he will not do is let organisations reach for the easy reading of it. AI is not simply going to absorb the junior work and leave the rest untouched. More likely, Dr Ang suggests, experienced employees will use it to take on tasks that once went to juniors – and a task that is peripheral to a senior’s role can be exactly the developmental work someone earlier in their career needs. Take that work away in the name of efficiency and the pipeline that produces senior people quietly thins, years before anyone notices the shortage.

His prescription is that junior roles stop being treated as a cost line to automate and start being treated as a workforce-planning question. Two responses are beginning to circulate. One is a revival of apprenticeship, structured deliberately around building judgement over time. The other is job redesign – rebuilding early-career roles around the tasks where human judgement matters most, such as the ambiguous cases and the work of reviewing and challenging what AI produces, so that juniors grow into those responsibilities under experienced guidance rather than being handed a role AI has already hollowed out.

That tension – between what is efficient now and what an organisation will need later – runs through everything Dr Ang has to say. And it starts with something far more mundane than succession planning: how HR measures whether any of this is working at all.

Stop counting completions

Dr Ang, a Principal at analytics firm Quaintitative who previously led the development of AI risk management frameworks at the Monetary Authority of Singapore, is blunt about the metric most learning functions still live by. Training completed, tools rolled out, courses finished – all of it measures activity, none of it measures whether anyone is producing better work. “AI fluency measurements need to be grounded in business outcomes and what success looks like for specific work tasks and workflows,” he says.

The reason that is harder than it sounds is that “good” refuses to hold still. Writing a report means one thing for an analyst and something else entirely for a customer service officer, so there is no single organisation-wide fluency score to report against. Which is why, Dr Ang says, HR and L&D “cannot do it alone.” The measure has to be built task by task, with the business leaders who own the work and can say what a good outcome actually looks like.

Done that way, the questions turn concrete. After training, did an employee use AI to produce work that meets the standard in less time? Do employees understand where AI fits across the steps of their own workflow? Can they weigh its speed against the moments that need human judgement, and evaluate what it hands back in the context of their own work? Answer those honestly and, over time, the organisation gets something worth more than a completion rate: a clear view of where in a workflow AI earns its place and where people still do. It is slower, messier, and near-impossible to reduce a dashboard figure. It is also the only version that describes what is actually happening.

The trap that looks like a win

If measurement is where good intentions go slack, cost-cutting is where they go wrong. The paper reserves a specific warning for what it calls the Extractive approach: using AI to strip out cost without building any capability alongside it. For an organisation under pressure to post short-term numbers, it is the path of least resistance – and Dr Ang is precise about why it is so tempting.

“The impact is asymmetric,” he says. Cutting headcount delivers savings that are immediate, certain and easy to report. Building capability asks for money up front against benefits that show up later and are harder to prove. Lay the two next to each other and the first wins almost every time.

READ MORE: Nearly eight in 10 employees in APAC use AI weekly. Only 7% of firms have scaled it

His case against it is that the ledger is missing entries. There is the illusion of productivity: surveys of desk employees find a large share regularly receive low-quality AI output that someone then has to fix, so individuals get faster while the organisations as a whole see nothing, because the workflows around them never changed. There is the cost of reversal – replacing people with AI has proved more complex than many organisations expected, a reminder that in customer-facing and judgement-heavy work the two are complements, not substitutes. And if employees come to see AI purely as a cost-cutting instrument, Dr Ang notes, they grow less willing to adopt it in the ways that would actually transform how work is done.

Then the quietest cost of all. “When capability leaves with the people who leave, the organisation becomes less able to respond to future change,” he says. The saving is visible on the day; the lost ability to adapt is not, until the moment the organisation needs to change and discovers it no longer can. Set the immediate efficiency against the slow erosion, and the Extractive route starts to look like what it is – a way to buy short-term savings at the price of long-term competitiveness.

Why waiting is the wrong move

The obvious objection is that all of this assumes a disruption that may not have arrived. On that, Dr Ang is careful not to overclaim. The experts the ATA interviewed genuinely disagreed on whether labour disruption is already underway or still anticipatory – but the split, he says, “largely comes down to what is being measured, rather than a genuine disagreement about what is happening.”

At the level of jobs – headcounts, hiring, wages – the change looks modest, and is hard to separate from the ordinary post-pandemic hiring slowdown. At the level of tasks, it is already measurable, and many organisations are living it. The two camps are reading the same reality through different instruments: one watching job-level data that moves slowly, the other watching task-level change that is happening now. Dr Ang adds a caveat HR should hold onto – most of the evidence so far is American, and Asia-Pacific is built differently, which is why he keeps returning to the need for research grounded in the region.

For an HR leader, though, the debate resolves into something most anticlimactic. Whether the job-level reckoning lands in one year or five, the task-level shift is already here, and the right response does not change with the timing. “Organisations can begin redesigning workflows, building AI fluency, and preparing their workforce now,” Dr Ang says, “rather than waiting for job-level changes to become more visible.”

This returns to the junior who used to learn by doing the dull work. The organisations that come out of this well will not be the ones that automate that work fastest. They will be the ones who noticed what it was quietly building and redesigned it before it was gone.