Which Skills Gain Value, Which Lose It
The Pattern Behind the Disruption
Technology has repeatedly automated the execution of a task while increasing the value of judging whether the output is any good. Spreadsheets did not eliminate accountants; they eliminated manual arithmetic and raised the value of financial judgement. AI follows the pattern with a wider reach, because it touches tasks previously considered safely cognitive: drafting, summarising, first-pass analysis, routine code. What tends to gain value is deciding what is worth doing, judging quality, and taking responsibility for the result. What tends to lose value is producing a competent first draft of something standard.
- Execution gets cheaper; judgement about what to execute and whether it is right gets dearer
- Producing a competent standard draft is no longer scarce, in text, code or images
- Deciding what is worth doing, and owning the outcome, remains hard to automate
- Whole jobs are rarely replaced; specific tasks inside them are, and roles reshape around that
Skills Worth Investing In
Judgement in a specific domain, because you cannot evaluate output in a field you do not understand. Genuine communication, meaning the ability to explain something to a particular person and change what they do, rather than merely producing prose. Working with other people, which includes disagreement, negotiation and trust. Anything physical and situational, which remains far harder to automate than office work. Original problem framing — noticing that everyone is answering the wrong question. And the compound skill of learning quickly, since the specific tools you master this year will not be the ones you use later.
- Deep domain knowledge, because supervision requires understanding
- Communication as persuasion and clarity, not as text production
- Interpersonal work: collaboration, negotiation, earning trust
- Problem framing — deciding which question is the right one to ask
The Awkward Bit About Entry-Level Work
Traditionally you built judgement by doing large amounts of routine work — the research memo, the basic analysis, the simple code, the first draft nobody wanted to write. That work taught you what good looks like. It is exactly the work most exposed to automation, which creates a genuine problem: how do you develop expert judgement if the apprenticeship tasks disappear? Nobody has fully solved this. What you can do is be deliberate. Do some work unassisted specifically to build the underlying skill. Review AI output critically rather than accepting it. Seek roles and projects where you are given real responsibility early.
- Routine junior tasks were how judgement was built, and they are the most automatable
- Compensate deliberately: practise unassisted, and critique output instead of accepting it
- Look for experiences with real responsibility rather than volume of routine tasks
- Being the person who can tell good from bad output is the fastest way to become useful
Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.