Keeping Your Own Judgement

The last warning in this book is the quiet one, and Tessa arrives at it by noticing two things in the same week.

The first is about herself. A supplier sends a difficult email, and her hand is on the chat box before she has finished reading it. Not because the reply is hard — because reaching for the model has become the first move rather than a considered one. Six months ago she read the email, thought for a minute, and then decided what to do about it. The minute has quietly gone.

The second is about a new colleague, and it is the more serious of the two. He is quick, capable, and has never written a summary of a supplier call without the model. When Tessa asks him whether a particular summary is any good, he cannot tell her. Not "he thinks it is fine" — he genuinely cannot tell, because he has no idea what a good one would have looked like.

Those two observations are the same problem at different stages, and this page is about the muscle underneath both.

Every Habit in This Book Assumes a Judge

Look back at what you have been taught to do, and notice what all of it rests on.

Spot the load-bearing fact and check it — that requires knowing which sentence carries the weight, which is a judgement made in about two seconds by someone who has written a hundred itineraries and not at all by someone who has written none. Read a random dozen rows against their reviews — that requires being able to tell a fair summary from a plausible one. Read the change before you keep it — that requires knowing what you were expecting to see.

Every one of those is a check, and every check runs on the same thing: a person who can tell good from bad. That is not a personality trait. It is a skill, built by doing the work enough times to know what right feels like — and like every skill, it fades when it is not used.

Which produces the awkward shape at the centre of this chapter. The tool works beautifully as long as somebody is checking it, and the checking works as long as somebody is still practised at the thing being checked. Remove the practice and the whole arrangement keeps looking identical from the outside, right up until the day something wrong goes through and nobody in the room notices.

This is not a warning about the model. It is arithmetic about the human half of the arrangement.

The Junior Problem

Tessa's checker's eye was built the slow way, by accident. She wrote summaries of supplier calls for years before there was any alternative — badly at first, then adequately, then well — and the hundreds she wrote are why she can look at a generated one and see instantly that it has dropped the only paragraph that mattered.

Her colleague will not get those years, because there is no reason for anyone to hand him work the model does in four seconds. The path that trained her has quietly closed behind her.

That is the honest version of the problem, and it does not have a comfortable answer. It has a deliberate one: what used to happen by accident now has to be arranged on purpose.

What that looks like at Waymark is not a training course. It is a small standing arrangement: for real tasks that matter, he writes the first version himself, and only then compares it with the model's. Sometimes his is better and he learns that he can already do this. Usually the model's is faster and smoother, and the comparison teaches him what smooth is worth and what it costs — which is exactly the lesson Chapter 3 spent a page on, learned in the only way it really sticks.

It costs him twenty minutes a week. It is the difference between a colleague who can check the output and one who can only pass it on.

Say plainly what this is: an apprenticeship, rebuilt on purpose because the old one has stopped happening by itself. Anyone senior enough to hand out work now owns that problem, whether or not they have noticed it.

Choosing What Stays Manual

The reflex response to all this is to do more by hand, generally, out of a vague sense of virtue. That is not a plan and it will not survive a busy month.

A plan is a written list, and Tessa's has two kinds of thing on it.

First, the tasks where her judgement is the product. The final read of anything that goes to a client. The pricing decision on a bespoke itinerary. Anything the card's first question rates as serious. These are not manual out of nostalgia; they are manual because what the client is actually paying for is her having looked at it.

Second, one regular repetition of the core skill. One supplier call a week summarized by hand, before the model sees it. It is not the efficient way to produce that summary and she does not pretend it is. It is the gym, not the workday — and like the gym, it only works if it is scheduled rather than intended.

Everything else can be delegated with a clear conscience, which is the point of writing the list rather than worrying about the question in general.

One reason to write it down rather than hold it as a good intention: drift is silent. Chapter 8 taught that about products — a model changes underneath, quality moves, and nobody notices until something breaks — and the same is true of people. Nobody decides to stop thinking first. It happens the way Tessa's missing minute happened, one convenient Tuesday at a time, and a list on paper is the only thing that ever catches it.

The Partnership, Stated Once

Consider what navigation did to drivers, because it is the clearest version of this that exists.

Turn-by-turn directions made every driver better the day they arrived, and nobody sensible wants to go back. What happened next is the part worth borrowing: the drivers who only ever followed the screen gradually lost the map they used to carry in their heads, and it shows on the day the bridge is closed and the screen is confidently routing everyone into a queue. The knowledge did not become useless. It became optional — and the good ones keep it deliberately, precisely because it is no longer forced on them.

Leave the drivers there. The point transfers cleanly enough: what a tool makes optional does not stop being valuable, it just stops being automatic.

So here is the arrangement this whole book has been describing, said once, plainly.

The model drafts, reshapes, sorts and explains, at a speed no person matches. You supply the context it cannot have, the standards it does not hold, the verification it cannot perform on itself, and the decision it is not accountable for. Every chapter since the first has been strengthening your side of that line — what the machine is, how to ask it, when to disbelieve it, what to keep out of it, what it costs, and when not to use it at all.

Keep that side strong. It is the half no meter can bill and no provider can ship, and it is the half that decides whether everything else in this book was worth learning.