# AI 101 · Lesson 3 — How to prompt

**The CV Method · AI 101 · Lesson 3 · Working with AI · Aug 2026**
*by Corina · @the_cv_method · hello.thecvmethod.com · current as of Aug 2026 — the AI space moves fast, so double-check tool names*

Good prompting isn't secret phrases you memorize — it's being clear about a few things. Get the structure right and your **first** draft is usually good, so you stop re-rolling the dice.

---

## The steps (at a glance)

1. **Role + goal** — the role it plays and the goal you want
2. **Guardrails** — length, tone, format, and what to avoid
3. **Example** — paste one 'this is good' sample to match
4. **No guessing** — give real inputs; make it ask or say 'unknown'
5. **Steer** — nudge the draft; keep what works

## Rule zero — Be clear, not clever.

You don't need magic words. You need to tell it who it is, what you want, and what 'good' looks like. Everything below is just that, in order.

## Step 1 · Say who it is + what you want

Start every prompt with two things: the **role** it should play and the **goal** you want. "You're my assistant — help me draft a reply that does X." That one line frames everything after it and stops the AI from guessing what you're actually after.

*The move: if the task is fuzzy, tell it to ask you questions before it answers.*

## Step 2 · Give the guardrails

Now the **constraints**: length, tone, format, and what to avoid. "Short, friendly, plain-English, as three bullets, no buzzwords." Boundaries are what turn a technically-correct answer into one you can actually use.

*The move: name the format you want — bullets, an email, a table. It changes the whole shape of the reply.*

## Step 3 · Show one example

Paste one example of "this is good." A single sample of the voice, length, or structure you want teaches it your standard far faster than describing it. This one trick fixes "it doesn't sound like me" better than anything else.

*The move: keep a few good examples handy to paste — your best email, your best caption.*

## Step 4 · Hand over your material — and ban guessing

Give it the actual raw material — the notes, the numbers, the doc — not a vague description. Then add the honesty rule: **if something's missing, ask or write 'unknown' — never make it up.** This is what keeps AI from confidently inventing facts.

*The move: "don't guess" is the single highest-leverage instruction for trustworthy output.*

## Step 5 · Steer, don't restart

When the draft is close, **steer it** — "keep all of that, just make the opening warmer." Don't rewrite the whole prompt from scratch. Small, specific corrections get you to great in two or three passes; starting over throws away the 90% that was already right.

*You've got it when: your first draft is usually close, and you fix it in a nudge or two, not a rewrite.*

## Keep it honest & safe

- Be clear, not clever
- Give it real context
- **"unknown"** beats a guess
- One nudge, not a restart
- **Date it** — AI moves fast

## Steal these prompts

**Role + goal**

```
You're my [assistant / analyst / marketer]. Help me [do X] for [who]. Before we start, ask me anything you need to do it well.
```

**Guardrails**

```
Keep it [short / friendly / plain-English]. Format it as [bullets / an email / a table]. Don't [overpromise, use buzzwords, invent details].
```

**Example**

```
Here's an example of the style I want: [paste a good sample]. Match this — same voice, length, and structure.
```

**No guessing**

```
Here's what I have: [paste your material]. If anything's missing, ask me or write 'unknown' — never make it up.
```

**Steer**

```
Keep all of that, but make it shorter and a bit warmer. Only change the opening line — leave the rest.
```