The skills that compound, and the ones that expire
Most AI advice optimises for this quarter. Here is how to tell which of your skills will still be worth something in 2028 — and which are already depreciating.
Every tool you learn has a half-life. The spreadsheet macro you mastered in 2015 is worth less today than the judgment you built while writing it. That asymmetry is the whole game.
Two kinds of skill
The distinction that matters is not technical versus non-technical. It is whether a skill compounds — whether what you learned last year makes this year’s learning faster — or whether it simply expires when the tool changes.
- Compounding: judgment about which problems are worth solving, taste in what “good” looks like, the ability to explain a decision to someone who disagrees with it
- Expiring: the specific syntax of a prompt, the layout of a settings panel, the pricing tier of a vendor that will be acquired next spring
Neither category is optional. You need working knowledge of the tools. The mistake is spending your whole learning budget on the expiring half.
A test you can run this week
Take the last five things you learned at work. For each, ask: if the tool disappeared tomorrow, what would remain? If the honest answer is “nothing,” you learned an interface, not a skill.
The people who will do well are not the ones who learned the most tools. They are the ones who built a way of working that survives the tools.
Where AI actually creates leverage
The leverage is rarely in the output. It is in the number of options you can afford to consider before committing.
- Generate more drafts than you could by hand
- Discard most of them quickly, on criteria you can articulate
- Spend the time you saved on the one decision that mattered
Step two is the one nobody teaches. It is also the one that does not expire.
Here is the shape of a prompt that supports step two — note that it asks for disagreement, not agreement:
Give me three approaches to <problem>.
For each: the strongest argument against it,
and what would have to be true for it to be the right call.Auditing your own role
A rough scoring exercise. Be honest rather than generous.
| Part of your role | Replaceable by a model? | Compounds over time? |
|---|---|---|
| Producing the first draft | Largely | No |
| Deciding what gets built | No | Yes |
| Explaining the trade-off | Partly | Yes |
| Maintaining the relationship | No | Yes |
If most of your week sits in the first row, that is not a verdict. It is a starting position — and it is information you can act on.