# Actually score your candidates

> **What you will learn**
> Not a chapter to read but one to **do**. Write your own candidates, score them
> against the three questions, and **throw half of them away.**

## What you need

```
□ A sheet of paper or a spreadsheet
□ 40 minutes
□ Ideally one person who does the work (alone, you fall into "they all look good")
```

---

## Step 1 — Write six candidates (10 min)

**Fill all six.** Do not stop at five. The one or two you have to force out are
often where the good candidate is.

Three places to look:

```
1. Where the time goes          → the widest block in the weekly schedule
2. What you decided not to automate → work that couldn't be reduced to rules then
3. What people put up with      → anything described as "that's just how it is"
```

| # | Candidate | Who does it | How often |
|---|---|---|---|
| 1 | | | |
| 2 | | | |
| 3 | | | |
| 4 | | | |
| 5 | | | |
| 6 | | | |

---

## Step 2 — Score loosely (5 min)

Mark ○△✗ against [the three questions](/guide/ax-choosing-tasks). Go fast and
by instinct for now.

| Candidate | Quality? | Cost? | Revenue? |
|---|---|---|---|
| 1 | | | |
| … | | | |

**Most people get a lot of ○ here.** That is expected — which is why step 3
exists.

---

## Step 3 — Score strictly (15 min)

Same candidates, harder questions. **If you cannot state the number, drop the ○
to ✗.**

| The loose question | The one to answer now |
|---|---|
| Does it cut cost? | **How many hours a month go into this today?** Where did you check that? |
| Does it improve quality? | **Quality measured how?** What is that value right now? |
| Does it help revenue? | **By what route does it become revenue?** How many steps in between? |

```
Scoring rule:
  ○ = I can state the number right now
  △ = I know where to look it up
  ✗ = I don't know / it isn't measured
```

**Writing ✗ for "I don't know" is the whole exercise.** Being generous here
costs you three weeks in Phase 2 later.

---

## Step 4 — Throw three away (10 min)

Move **at least three** of the six to shelved. This is a rule, not a
preference.

| Verdict | Criterion |
|---|---|
| **Candidate** | at least one ○ under the strict questions |
| Shelved | everything △ or below |

Two things to watch while cutting.

### If "AI training for staff" survived

**That is an activity, not a project.** It scores ○ on none of the three
questions. It may well be necessary — it is not a first project.

### If the biggest, most important thing ranked first

**Be suspicious.** A first project is chosen for **speed of proof**, not size.
Pick something that cannot show a number within three months and attention will
cool before it does.

---

## Self-check

```
□ I filled all six candidates
□ At least one ✗ appeared under the strict questions   ← if not, you went easy on them
□ I moved three or more to shelved
□ I picked among the rest by "will this show a number within three months?"
```

If the second line is unchecked, go back to step 3.

---

## Check yourself

**1. Why insist on six candidates?**

<details>
<summary>Answer</summary>

**Because the ones you have to force out are often the good ones.** The three
that come easily are the ones everyone already knows — usually already tried, or
not done for a reason.
</details>

**2. Why score loosely and then strictly?**

<details>
<summary>Answer</summary>

**Because the loose question returns ○ for everything.** Everyone says yes to
"does it cut cost?". Change it to "how many hours a month, and where did you
check?" and most drop to △ or ✗ — which is when a ranking finally appears.
</details>

**3. Why not start with the biggest, most important project?**

<details>
<summary>Answer</summary>

**A first project is chosen for speed of proof, not size.** Without a number
inside three months, organisational attention and budget cool off. Big projects
come after the first success.
</details>

---

You know what to do. The next part is **the procedure for actually running it**
→ [The roadmap](/guide/ax-roadmap-overview)