ACAI Booster Shot #3: Deskilled or Never-Skilled? Why AI Is Magic for Some and a Crutch for Others

This week I read a piece in The Straits Times by Rachael Bedard, a geriatrician and palliative care doctor, first published in The New York Times. She confesses that AI has made her a better doctor. She uses it several times a day, and it often reminds her of possibilities she had missed. But she also worries that the same tool is making her trainee doctors worse.

It raised a paradox I have been thinking about for a while: why is AI like magic for some professionals, but a crutch for others?

The difference is what you bring to it

Bedard has twenty years of practice behind her. When she reads an AI answer, she checks it against a mental model she has built case by case. She knows what a good answer looks like, so she can tell when something is off.

Her students do not have that yet. They have memorised less, and they tend to accept what the screen says because they have nothing to compare it with.

Same tool. Very different outcomes.

Where my own judgement came from

Reading this took me back to my articleship days. Calculating advance tax manually, filling income tax forms by hand, working out capital gains. It was slow, tedious work, and at the time I did not see the point.

Years later, I understood. That grind is where I learnt what a wrong number looks like. When a figure seems off today, I often sense it before I can explain why. No shortcut could have given me that instinct.

The words that stayed with me

Bedard quotes Adam Rodman, a physician and AI researcher at Harvard Medical School. For an experienced doctor, he says, relying on AI may cause a little deskilling. For someone still learning, the risk is "never-skilling."

Never-skilled. Those words caught my attention, and I believe this is a real fear across industries.

Deskilling is when you lose a skill you once had. You can rebuild it. Never-skilling is when the skill was never built in the first place. You do not even know it is missing.

Think of the young auditor who never learns to trace an entry back to source. The junior lawyer whose contract review is only as good as the AI's first pass. The new coder who can ship a feature but cannot debug it. The junior analyst who can generate an automated forecast but cannot tell if the cash flow logic makes sense.

Someone who uses AI before acquiring the basics faces three problems:

• They will not know what is missing. A neat, confident answer can still have gaps.

• They will not spot hallucinations. A wrong answer delivered confidently looks exactly like a right one.

• They cannot point it in the right direction. Good output needs good questions, and good questions need understanding.

The MSG test

Think of MSG, the flavour enhancer. Tasted on its own, it offers very little. Added to a well-made dish, it makes the flavours come alive.

Add it to a dish that lacks good ingredients, however, and you only end up masking the absence of real flavour.

AI is the MSG of knowledge work. It enhances what is already there. If the professional has depth, experience, and judgement, AI makes the output richer. If the foundation is missing, AI cannot supply it.

A message for leaders

If never-skilling shows up anywhere first, it will be in your junior pipeline. Their work will look good. The gaps will surface years later, when they are the ones expected to review, decide, and lead.

Rolling out AI tools will not solve this. Leaders have to design the struggle back into how people learn:

• Build before you boost. Let new joiners work through real problems on their own first, then compare their thinking with the AI's.

• Make seniors visible. Ask experienced people to show how they question and verify AI output. Judgement is easier to learn when you can watch it in action.

• Review the reasoning. If someone cannot explain how they reached an answer, they have not learnt it yet.

• Reward the catch. Recognise the person who spots what is wrong, alongside the one who delivers quickly.

For boards, there is an often overlooked governance question: how is AI shaping our talent pipeline? A generation that never develops independent judgement is a serious succession risk, and one that will not appear on any risk register until it is already institutionalised.

Mastering the fundamentals first

As Bedard says, the hardest parts of any profession will stay hard. AI will make many things easier, but only for those who have first mastered the fundamentals.

Get the dish right first. Then add the MSG.