Name the skills in your work most likely to fade if AI does them for you every time.
When an AI tool does a task for you every day, you stop doing that task yourself. The skill behind it then weakens, often without you noticing, until the day you need it and find it slower or harder than it used to be. This weakening is called skill decay, and it can be prevented if you know which skills are exposed and keep using them on purpose.
Many people can still recite phone numbers they memorised as children: a friend's number, their parents' home line, the number of their first office. Ask them for a number they dialled fifteen years ago and it often comes straight back. Ask for their partner's current mobile number and many cannot give it. Their memory did not fail. The phone stored the number, so they stopped needing the skill, stopped practising it, and it faded. The same fading can happen to more important skills once AI does them for you daily.
Skills need practice to stay sharp. This is an old idea, and it has nothing in particular to do with AI. People who stop driving for a few years often feel nervous behind the wheel when they start again. People who once wrote long reports from scratch find a blank page harder after years of filling in templates.
AI speeds this up and affects more skills at once, because it can now do many tasks that used to need practice. The skills most often mentioned as at risk are:
writing from a blank page mental arithmetic and rough estimation reading a long document carefully summarising an argument in your own words structuring a presentation working through a problem step by step before reaching for an answer
If the AI does a task well, keeping the skill yourself might look like nostalgia. The problem is that the skill you stop practising is often the same skill you need to check the AI's work.
Lesson 7.1, "Why we trust confident machines too much", showed that automation bias is strongest when you could not easily do the task yourself. Skill decay moves more tasks into that group. Siew Ling, the accountant from that lesson, spent months asking an AI assistant to draft her variance commentary. During a system outage she sat down to write one by hand and found it surprisingly slow. She knew what she wanted to say, but the words did not come as easily as before. If her ability to read a variance table and write about it fades further, so does her ability to spot when the AI's commentary on that table is wrong.
Other skills work the same way. If your mental arithmetic gets weaker, you are less likely to notice a spreadsheet formula or an AI calculation that is off by a factor of ten. If you stop reading long documents closely, you lose the sense that tells you a summary has left something important out. Less practice means less ability to catch errors, and that means you depend more on the tool being right.
For experienced people like Siew Ling, the risk is that a skill fades. For people early in their careers, the risk is that the skill never forms in the first place.
Ravi is a graduate who joined Siew Ling's team this year. Since his first week he has used an assistant to draft almost everything, and his output looks polished. But he has not yet written fifty variance commentaries by hand, struggled with them, and had them marked up by a senior colleague. That cycle of struggling with the work and getting it corrected is how most professionals develop judgement. Without it, Ravi may produce good-looking work without understanding why it is good, or without noticing when it is not.
None of this means beginners should avoid AI. It means beginners and their managers should decide which skills are built by hand first. Ravi's manager asked him to write his first three commentaries by hand before using AI for them.
The fix is to keep practising the skills you care about, on purpose, even when AI could do them. Think of these as reps, the deliberate repetitions that keep a skill in working order. None of the following habits require you to give up what AI does for you.
Draft first, then ask the AI to critique your draft. You do the hard thinking and the assistant improves it, instead of the other way round.
Estimate before you calculate. Make a rough guess in your head, then check it against the tool's result. When the two are far apart, you have a reason to look more closely.
Read the original before the summary, at least for the documents that matter. You will then know whether the summary has dropped anything important.
Pick one task a week and do it entirely by hand, the way an athlete keeps doing basic drills long after mastering them.
Siew Ling now writes the first paragraph of each month's commentary herself and has the assistant draft the rest. Ravi is writing his first three commentaries by hand. Neither change takes much time, and both keep the underlying skill alive.
Now look at the tasks you hand to AI most often. Behind each one is a skill you used to practise, whether writing, calculating, reading, structuring or deciding. Note which of those skills you still use yourself.
List five tasks you now give to AI. For each, write the skill it depends on and whether you still practise that skill.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).