# Design a POC on one page
> **What you will learn**
> Not a chapter to read but one to **do**. You will design one POC on **a single
> page**. The most important box is not the success criteria — it is the **stop
> conditions.**
## What you need
```
□ 40 minutes
□ One candidate task chosen earlier
□ One person who actually does that work — without them, do not start
□ One page or a spreadsheet
```
If you have not picked a task yet, do [scoring your candidates for
real](/guide/ax-try-scoring) first.
**That third line is a condition, not a suggestion.** A POC designed without the
people who do the work stalls three weeks in, at "that is not how we actually do
it."
---
## Step 1 — Narrow it to one sentence (5 min)
```
✗ "Improve customer enquiry handling with AI" <- no boundary
✓ "Draft the first reply to return enquiries" <- one kind, one stage
```
### What to look for
```
□ Is it one kind of work? (all enquiries ✗ / return enquiries ✓)
□ Is it one stage? (the whole response ✗ / first draft ✓)
□ Can you name the output in one word? (a draft)
```
**If any answer is no, it is still too wide.** The purpose of a POC is not
results but **a decision** — the narrower it is, the sooner you decide.
---
## Step 2 — Write today's numbers (10 min)
If you already have a [baseline](/guide/ax-try-baseline), copy it across.
Otherwise write only what you can answer right now.
| Item | Today's value | Where you checked |
|---|---|---|
| Volume (week/month) | | |
| Time per case | | |
| What quality is measured by today | | |
```
Blank rule:
If you cannot state a value, write "unknown" in the box.
Two or more "unknown" means: do not start a POC. Spend two weeks measuring.
```
**Write "unknown" rather than leaving it empty.** An empty box looks like
something you forgot; "unknown" is a decision.
---
## Step 3 — Success criteria (10 min)
It is only a success criterion if it contains **both a number and a date.**
```
✗ "Quality improves"
✗ "The team is happy with it"
✓ "In six weeks, 60% or more of drafts are sendable as written"
✓ "In six weeks, average handling time falls from 12 min to 7 min or less"
```
| Our criteria | When | What | How much |
|---|---|---|---|
| 1 | | | |
| 2 | | | |
**Two at most.** Write more than three and you will hit one of them and call it a
"partial success."
---
## Step 4 — Stop conditions (10 min)
**This is the point of the exercise.** Most POC documents have no such box, which
is why POCs that are not working run for six months.
```
On hitting any of these, stop immediately and re-decide:
□ At week 3, if ( ) is below ( )
□ If review time exceeds ( )% of the original handling time
□ If the people doing the work say "it is more hassle" two weeks running
□ If a single irreversible incident occurs -> stop at once
```
### Why this matters more than the success criteria
Success criteria are used **when it works.** Stop conditions are used **when it
does not** — and most POCs land on the second side. With no basis for that
judgement written down, **"let's give it a bit longer" repeats**, because nobody
finds it easy to propose killing the thing they started.
> **Stop conditions can only be written honestly before you start.** Afterwards,
> the money already spent bends the judgement.
---
## Step 5 — Collect it onto one page (5 min)
```
[POC name]
[One sentence] what, and up to which stage
[Duration] start ~ end (six weeks recommended)
[Today's values] volume / time / quality (including "unknown")
[Success] two or fewer, with numbers and dates
[Stop] four conditions
[People] practitioner ( ) delivery ( ) decision-maker ( )
[Out of scope] three things you are not doing this time
```
**Fill in that last line.** Without "out of scope" written down, the scope grows
mid-flight, the growth breaks the deadline, and the broken deadline postpones the
decision itself.
---
## Self-check
```
□ The sentence is narrowed to one kind of work and one stage
□ You wrote "unknown" honestly where you did not know
□ Success criteria number two or fewer, with a number and a date
□ You wrote four stop conditions <- if this is empty, the exercise failed
□ The decision-maker is named
```
If the fourth is unchecked, go back to step 4. **A POC without stop conditions is
not a design — it is a request for approval to begin.**
---
## Check yourself
**1. Why do stop conditions matter more than success criteria?**
Answer
**Because most POCs end up on the "did not work" side.** Success criteria are only
used when it works; without a written basis for the other case, "a bit longer"
repeats. And after starting, sunk cost makes honest writing much harder.
**2. Why write "unknown" instead of leaving the box empty?**
Answer
**"Unknown" is a decision; an empty box looks like an oversight.** Two or more of
them immediately yields the judgement that you should be measuring, not running a
POC.
**3. Why write down what is out of scope?**
Answer
**Because scope grows during the run.** The growth pushes the deadline, and a
pushed deadline removes the moment of decision entirely. Nailing the boundary
before you start is cheaper.
---
The final stage → [Phase 4 — production rollout](/guide/ax-phase4-rollout)