Enabled User · Module 01

The First Word Matters

After this, you can choose a verb that matches what you actually want, and predict how much the AI will decide to do on its own based on that choice.

Entering Stage 2 · Enabled User

You can use AI. Now learn to control it.

Intro

In B-10 you learned to tell the AI when to stop. The verb you use determines where it starts. One word sets the ceiling on everything that follows.

The concept

Send the AI "Improve this email" and it will decide what improvement means. It might shorten it. It might restructure it. It might change your tone, cut your examples, add a new section. Every decision it makes feels justified — it was asked to improve.

Send "Rewrite this email to be more direct. Keep it under 150 words. Don't change the opening." and it has a job with a defined edge. It knows what to do and where to stop.

The difference is not the task. It is the first word.

Some verbs have a clear ceiling — the AI does the thing you asked and stops. Others have no ceiling at all — the AI decides what "better" means, and then decides what counts as better, and then decides how far to go. When you use a risky verb, you are not asking for a result. You are delegating a judgement.

The exercise

Part 1: Side by side

Pick something from your own work — an email, a paragraph, a message you have been meaning to send. Something short. Something real.

Now send two versions to your AI tool. Use the same piece of work for both.

Version A (risky verb):

Improve this [email / paragraph / message].

[Paste your text here.]

Version B (safe verb + stopping condition):

Rewrite this [email / paragraph / message] to be [one specific change: more concise / clearer / more direct]. Keep it under [word count]. Don't change [something you want left alone].

[Paste the same text here.]

Copy these starters. Replace the bracketed parts with your own task. Send both.

Then compare:

What did Version A change that you did not ask it to change?
Did Version B stay inside the limits you set?
Which response would you actually use?

You are not looking for a "better" response. You are looking for the difference — what the first word gave the AI permission to do.

Part 2: A real rewrite

Think back to the past week. Is there a message you sent to the AI where it did more than you wanted — rewrote too much, changed your voice, added things you didn't ask for?

If you have one, try it again now. Rewrite that message using a safe verb. Use the full pattern:

[Safe verb] [the specific thing]. [Stopping condition]. [What to leave alone].

Send it. Compare it to what you got before.

If nothing comes to mind, use the same piece of work from Part 1 and try two more verb swaps: once with "Fix" (minimal change), once with "Summarise" (condense, don't rewrite). See what each one does.

Your verb reference

The full verb reference for this course is at:

/course/reference/verbs

Keep it. You will use it in every module from here.

It covers seven safe verbs — Summarise, Explain, List, Fix, Draft, Rewrite, Compare — with worked examples and a "How to edit this" guide for each one. It also covers the risky verbs (Improve, Clean up, Make better, Enhance, Optimise) and what to use instead of each.

When you are not sure which verb to reach for, open that document first.

What good looks like

The response does exactly what you asked and stops. You can see where the edge is — the AI stayed inside it.

If you are not sure which verb to use

Open the verb reference and read the "What the AI will do" and "What the AI will not do" lines for the verbs you are considering. The one where "What the AI will not do" matches what you want left alone is the right verb. If none of the safe verbs fit your task precisely, pick the closest one and add a constraint: "and don't change anything else."

Now that you can choose a verb that sets the right ceiling, E-02 shows you how to protect the specific parts of your work you want left alone — before the AI starts, not after.