Cognitive Transition · Book One

Chapter 9 · Real Work in Transition

“An output with no judgment behind it is a risk with good spelling.”

Summary

A headline claims that AI now does a lawyer's job. Underneath the headline sits a task, a draft the model writes in seconds, and around it a whole system that is still in a person's hands: the client who asks for one thing and needs another, and a signature with a bar license behind it. This chapter brings the distinction between automating tasks and automating systems down to the ground of employment. It explains why the boundary moves with every generation of models without ever disappearing, what changes when you are paid for deciding more than for doing, and how to tell the organization that bets on judgment from the one that saves money on it. It also looks at education, and at the professional with years in the trade, who tends to underestimate how scarce their experience has become. It closes with a tool, the 5-Year Learning Investment Decision, for aiming your training where the value stays.

Extended content

Liquid part: this text keeps growing after the printed edition.

Twenty minutes with your job description

Open the document that says what you do. If there isn't one, write it from memory. Underline every verb and sort the verbs into two columns: "draft," "sort," "prepare," "summarize" on the left; "decide," "negotiate," "prioritize," "answer for" on the right. Tasks live on the left. The system lives on the right.

With that sheet beside you, apply the read-it-twice test for a week. Every headline about your sector that crosses your path, read it again and ask which task is underneath and which column it belongs to. Usually the headline is talking about the left, while your paycheck depends on the right.

Then comes the uncomfortable pass: count the hours of your real week that land in each column. If the right-hand side comes out nearly empty, your job isn't expendable. What's happening is that your calendar hides the system behind the production.

A worked example

Inés has spent almost two decades in purchasing at a mid-sized construction firm. When someone asks what she does, she says, "I negotiate with suppliers and prepare the orders." Half of that sentence is pure task, and a model can already do that half with reasonable supervision.

The other half isn't in the sentence. Inés knows which supplier runs late precisely when it rains, who shouldn't be pushed in October, and which cheap quote hides an impossible deadline. None of it appears in her job description, and so nobody protects it, herself included.

What she does is write three one-page cases about decisions of hers. In the first she describes an order she advised against signing, and what signing it would have cost. When her manager reads it, he stops asking how many orders she processes a month and starts asking what she thinks about the next ones. That shift is the real result of the exercise.

A portfolio case, in four fields

Try it with a decision of your own from the past year. Write it in about a hundred and fifty words:

Context: what was at stake and who expected what.
Alternatives: the real options, including doing nothing.
Decision and reasoning: which you chose and why that one and not another.
What it prevented, or what it cost: the consequence, even if the consequence is that nothing happened.

The fourth field is the hard one, because the best cases are the ones where nothing happened and you had to know why.

Taking the three conditions to your manager

Time, recognition, autonomy: the chapter teaches you to measure them. What's left is how to ask for them, and an abstract complaint gets filed away. A limited trial works better: for one month, a protected weekly block for reviewing outputs, logged as a task and not as a favor, and the authority to send back a deficient one without asking permission. At the end of the month, you bring what was caught. A trial with results gets discussed, while a vague request for recognition gets put off.

If someone reports to you, run the same calculation on their week. Do they learn from work the model half did, or do they also go through the complete job?

Questions to work with

Which part of my week is task and which is system, and which one do I mean when someone asks what I do?
Where has the line moved in my trade this year, and what happened to whoever stood just beneath it?
What do I know about my work that I have never written down?
What signs of seniorization or juniorization do I see in my own organization, and which ones do I keep quiet about?
Which decision this year was good precisely because nothing happened?

Where to next

Knowing where your value lies is the first step. The second is governing your own use of AI so that value doesn't erode without your noticing, and that is the work of Chapter 10.

Difunde el mensaje