Running the method on your own trade

Read the next lessons even if yours is not there

The lessons after this one are written for particular trades — accounts, teaching, health, law, government, shopkeeping, marketing, the practical trades. Yours may not be among them, and that is fine, because they are all the same five questions applied to different material. This lesson is the method, so that you can run it on your own work and get the same answer a specialist lesson would have given you.

The five questions

1. What do I produce in volume, where the same judgement recurs?

Volume is where the value is, because a saving of six minutes matters when it happens ninety times and does not when it happens twice. Look for the thing you do weekly that has a shape: the visit report, the quotation, the case note, the shift handover, the parent letter, the grant update, the inspection record.

2. Where does the truth live for this output?

Four possible answers, and they determine everything:

• In my head — my judgement, my observation, my decision. The model can format it. It cannot supply it.
• In a document I hold — attach the document, and the failure mode drops from invention to misreading.
• In a system — the CRM, the ledger, the patient record. Get the data out first; do not ask the model to remember it.
• In the world — the current regulation, the market price, what happened last week. This is the expensive case, and it is where fabricated confidence lives.

3. Who is harmed if it is wrong, and can it be undone?

An internal draft nobody acts on: harmless. A quotation sent to a customer: recoverable, embarrassing. A dose, a filing deadline, a safeguarding referral, a structural calculation: not recoverable. The irreversibility is what sets how much checking is proportionate, and it is a property of the output, not of the tool.

4. Is there a regulator, a registration or a professional body with a view?

If you hold a registration you can lose, somebody has published guidance. Find the actual document — your institute, your council, your bar, your board — and read it rather than a summary of it. Most of it is short, most of it is sensible, and it is what you will be measured against.

5. Can I check the output faster than I could have produced it?

The question from earlier in the course, and the one that decides. If checking costs more than doing, the task is not a candidate however good the draft looks.

The two tests that override everything

The decision-about-a-person test. If the output is a decision affecting somebody's rights, money, health, liberty, education or livelihood — hiring, dismissal, grading, benefits, credit, admission, discipline, care — a different set of rules applies, in most jurisdictions and in every professional code. That is its own lesson in the previous block, and the answer is never "the model decided".

The who-signs test. Whoever signs it is accountable for all of it. Not for the parts they wrote; for all of it. If your name goes on the bottom, the model's contribution is your contribution, and every honest professional standard treats it that way.

The pattern transferring

Four quick examples, to show that the trade matters less than the shape.

A plumber quoting jobs: the volume task is the written quotation. The truth lives in his head, from the survey. The model formats and explains, and never invents a price.

A translator: the volume task is a first pass. The truth is in the source text — attached. The risk is a fluent, wrong rendering of a term of art, so she keeps a glossary and checks it every time. Irreversible if it is a contract.

A librarian writing reading lists: the truth is in a catalogue, which is a system. Get the records out, then let the model write the annotations from them, never from memory — a fabricated ISBN is the classic failure.

A cooperative secretary writing minutes: the truth is in the meeting. Record with consent, transcribe locally, structure the transcript, and check every figure and every name against what was said.

Same five questions. Four different answers.

Do it now, on paper

Take the last full week. List every recurring output. Against each, write the four-way answer to question two and a one-word answer to question three. Then circle only the rows where the truth is in a document you hold or in your own head, and where being wrong is recoverable.

That circled set is where to start, and it is usually two or three things rather than twenty. Everything else either needs the sourcing discipline from the earlier blocks, or belongs in the column marked never.

Nothing in the lessons that follow is advice for your particular situation, and regulation varies by country and changes without notice. They are worked examples of this method, by people who have to live with the consequences.