When production gets cheap, something else gets expensive
For most of working life, being able to produce things was the valuable part. Writing the report, building the spreadsheet, drafting the plan, making the deck. If you could produce, you had a job.
That is the part getting cheaper. Not free, and not perfect, but cheap enough that being the person who can produce a competent first draft is no longer the rare thing it was.
Here is the part people miss. When production gets cheap, the bottleneck moves. It does not disappear, it relocates. And it relocates to the question of which output was any good, which one was worth making at all, and who can tell the difference. The scarce skill stops being production and becomes discrimination.
In plain English
- Discrimination:
- Used here in its older sense: the ability to tell good from bad, and better from worse. Nothing to do with the modern legal meaning.
- Domain knowledge:
- Deep familiarity with a particular field, industry or organisation. The kind of thing that takes years and lives partly in your instincts.
- Plausible-wrong:
- Output that reads well, sounds confident, and is incorrect. The most dangerous failure mode, because nothing about it looks broken.
- Taste:
- A reliable, quick sense of which version is better, built from seeing many examples and their consequences.
Four skills that go up in value
1. Judgement about what is worth doing
The cheapest thing in the world now is a competent piece of work nobody needed. When output is easy, teams drown in it: more reports, more analyses, more decks, more variants. The person who can look at ten possible things and say that only two matter, and defend that choice, becomes the constraint everyone routes around.
This is not a soft skill. It is the ability to hold the actual goal in mind and refuse work that does not serve it.
2. Domain knowledge deep enough to spot a wrong answer
You cannot check what you do not understand. An AI answer about your industry will read fluently to an outsider whether it is right or catastrophically wrong. Only someone who knows the field can feel the wrongness, and often they feel it before they can explain it.
This is the strongest argument against the instinct to chase tools. The tools change constantly. What lets you use them safely is depth in something specific, and depth compounds.
3. Clear writing about ambiguous problems
Not writing that sounds nice. Writing that takes a messy, contested situation and states plainly what the problem is, what the options are, and what we do not know yet. AI is genuinely good at producing polished prose about a well-defined problem. It is much weaker at deciding what the problem actually is when nobody has said so yet.
4. Explaining a decision
Producing an answer and being able to defend an answer are different jobs. The second one requires you to know why, to have considered what you rejected, and to hold up under questions from someone who disagrees. That is what a room full of people is really testing when they push back.
Notice what these four have in common. Every one of them is about the decision around the work rather than the production of the work. That is where the value is moving.
What this means practically
The instinct when you feel behind is to learn tools. It feels productive, there are courses everywhere, and it is measurable. It is also the least durable thing you can do, because the specific tool you learn will change under you.
The durable version is to get better at the four skills above, using AI as the training environment rather than the subject.
โ Weak prompt
Prompt
Write me a recommendation on whether we should switch suppliers.
Output
Switching suppliers could offer cost savings and improved service levels, though it carries transition risk. A phased approach is often advisable.
You have outsourced the judgement and got back a shape with no content. You cannot defend a word of this in a meeting.
โ Good prompt
Prompt
I think we should switch suppliers, for these three reasons. Here is what I know and here is what I am unsure about. Ask me the six questions a sceptical finance director would ask, hardest first. Do not write the recommendation.
Output
1. What is the true cost of transition including staff time? 2. What happens to the contract we are 8 months into? 3. Who told you the new supplier's lead times, and have you verified them?
The judgement stays yours. The AI is used to stress-test it, which is exactly the thing that makes your final answer defensible.
Here is a decision I am about to make at work, with my reasoning.
Do not tell me what to do. Instead, ask me the eight questions a well-informed sceptic in my organisation would ask, ordered from most uncomfortable to least. After each question, add one sentence on why it is the question that would expose a weak argument.
MY DECISION AND REASONING:
[paste]
I have written up a messy situation at work below. Help me sharpen the problem statement, not the prose.
Give me: (1) the clearest one-sentence statement of what the actual problem is, (2) two alternative framings of the same situation that would lead to different decisions, and (3) a list of the things we genuinely do not know yet.
THE SITUATION:
[paste]
There is a trap in all of this. If you use AI to do the judging as well as the producing, none of these skills develop, and you end up unable to check the thing you are signing off. That is the subject of lesson four, and it is the single most important idea in this course.
Checkpoint
When producing work gets cheap, deciding what is worth producing and whether it is any good becomes the scarce skill. Judgement, depth in a domain, clear writing about ambiguity, and defending a decision all go up in value.
A way to practise this week
Take one piece of work you would normally just do. Before starting, write three sentences: what the point of it is, who it is for, and what would make it good. Then produce it however you like, with or without AI. Then check the result against your three sentences.
Here are three sentences I wrote before starting a piece of work: what the point was, who it was for, and what would make it good. Below that is the finished work.
Judge the work only against my three sentences. Tell me where it drifted, where it is padded, and what a reader who cares about my stated point would skip. Do not rewrite it.
MY THREE SENTENCES:
[paste]
THE WORK:
[paste]
Do that a dozen times and you will have built something no tool update can take away: a reliable sense of what good looks like in your own field.
๐ Quiz
Question 1 of 4As producing work becomes cheaper, what becomes the scarce skill?