The AI paradox: make your job obsolete

In 2005, an online chess site ran a tournament with no restrictions on who — or what — could enter. Grandmasters entered. So did the strongest chess engines. So did teams of both. The winners were a pair of American amateurs running three computers at once.

Garry Kasparov, who had lost to IBM’s Deep Blue eight years earlier, drew the lesson himself:

“Weak human + machine + better process was superior to a strong computer alone and, more remarkably, superior to a strong human + machine + inferior process.”

Read that last clause again. The stronger player lost to the weaker one. What separated them was not chess ability. It was the quality of their process for managing the machine.

Twenty years on, that is the sentence every organisation is living inside. And it creates an uncomfortable problem: the better your people become at managing the machine, the fewer of them you appear to need.

The paradox nobody says out loud

The employee who genuinely masters AI does not simply work faster. They redesign the work. They discover which parts of their role were never really judgement at all, and hand those parts to a system that completes them in seconds. In doing so, they produce something no consultant’s report ever could: irrefutable evidence that the role can be done by fewer people.

We are asking people to volunteer that evidence. We have not told them what they get in return.

They have noticed. In a study of more than 48,000 people across 47 countries by KPMG and the University of Melbourne, 57% of employees said they hide their use of AI at work, and almost half admitted using it in ways that contravene company policy. This is not a technology adoption problem. It is a trust problem wearing a technology costume.

And the concealment is rational. If mastery might shrink your function, the sensible move is to use AI brilliantly, keep the time you save, and say nothing. Every organisation congratulating itself on adoption rates should ask how much of its real productivity gain is currently being banked privately — and what that tells us about what people expect from us.

Why ‘your job is safe’ is the wrong promise

The instinctive leadership response is reassurance: AI will augment, not replace. Nobody’s job is at risk.

It is the wrong promise, for two reasons. First, it is frequently untrue, and people can do the arithmetic themselves. If a function of 40 becomes 30% more productive and the workload does not grow, the honest answer is not “we will all simply do more”. Second, and more importantly, it defends the wrong thing.

Job security has always rested on scarcity — being hard to replace because what you did was hard to do. That is a bet on the durability of a set of tasks. The World Economic Forum (WEF) expects nearly 40% of the skills needed for today’s jobs to change by 2030, with 92 million roles displaced, 170 million created, and more than 120 million workers whose prospects are at risk because they are unlikely to receive the reskilling they need.

Against numbers like those, protecting the shape of your current role means protecting a depreciating asset. You can win that argument for three years and still lose your career.

What actually appreciates

The better conversation is not about security. It is about value — and about which of your capabilities travel with you. Three of them do.

Framing: AI answers questions with startling fluency; it has no view on which question is worth asking. Deciding what the business should solve this quarter, and what it should stop doing altogether, never gets automated, because it is not a task. It is a judgement about value.

Verification: The WEF study found that 66% of people rely on AI output without evaluating its accuracy. Production has become almost free; being wrong has become expensive. The scarce person is no longer the one who can generate the analysis, but the one who can look at twelve versions of it, identify which is right, and put their name to it. Accountability does not scale down. Someone still has to sign.

Mobilising others: Most AI value in large organisations is not lost in the model. It is lost in the last mile — stranded between a working prototype and a changed way of working, because nobody could persuade forty people to abandon a process they were already good at. The technology arrives in weeks; the behaviour takes quarters.

None of these three belong to a job title. They belong to the person, and they travel — across roles, across functions, across employers. Which is why the person who makes their own job obsolete usually turns out to be the safest one in the building: they now understand the work well enough to redesign it, and everyone around them knows it.

Security came from being hard to replace. Value comes from being easy to redeploy. The first is a wall; the second is a passport.

Dr Philios Andreou Sphika is the Deputy CEO and President of the Other Markets Unit, including Thailand and Asia, at BTS Group, a leading global strategy execution consulting firm specialising in the people side of strategy. He is deeply committed to working closely with teams and businesses to drive meaningful impact on a global scale. For executives who are interested in connecting, Dr Philios can be reached at philios.andreou@bts.com or visit his LinkedIn profile.