# Delegate · Describe · Discern · Take responsibility
> **What you will learn**
> You have been through [your first try](/guide/ai-first-try) and [getting the
> answer you wanted](/guide/ai-how-to-ask). This chapter is the map that shows
> **where those sit in the whole.** There are four moves in working with AI, and
> the difference between people who use it well and badly is usually not the
> prompt but the other three.
## "Just write good prompts" is not even half of it
Almost everyone who sets out to learn AI starts hunting for prompt tricks. But
collect the incidents that actually happen in the field and the order reverses.
```
Cause of incident Which move failed
──────────────────────────────── ─────────────────────
Handed over work that never should be Delegate
Said it vaguely Describe
Used the answer as it came Discern ← most common
Something wrong shipped, nobody knew Take responsibility
```
**The prompt (Describe) is one move out of four.** And real incidents mostly come
from the third and fourth.
```mermaid
graph TD
A["1. Delegate
What to hand over"] --> B["2. Describe
How to ask for it"]
B --> C["3. Discern
How to look at what came back"]
C -->|"describe again"| B
C --> D["4. Take responsibility
What happens after it ships"]
```
## 1. Delegate — what to hand over
**The decisions you make before asking.** Is this work suited to AI at all; do I
do it with AI assisting, or do I hand the whole thing over?
There are three ways to hand over. Naming them makes meetings tractable.
| Mode | Meaning | Example |
|---|---|---|
| **Automation** | Runs to the end with nobody watching | Sorting inbound enquiries |
| **Augmentation** | I do it, AI assists | I take a draft and revise it |
| **Agency** | Give the goal, leave the method | "Gather this material and write it up" |
**The same task carries different risk depending on which of the three you
choose.** Contract review as augmentation means a person looks every time. As
automation, nobody looks. Same work, different decision.
```
✗ "Let's automate contract review with AI"
✓ "Extracting risky clauses is automatic; the judgement is a person's"
```
- Choosing the work → [Three questions for choosing work](/guide/ax-choosing-tasks)
- How far to hand over → [Deciding how much to hand over](/guide/ht-autonomy)
## 2. Describe — how to ask for it
**Getting what is in your head out into the open.** The model does not know what
you know — your company, this document's readers, what was decided in yesterday's
meeting.
```
✗ "Tidy up this report"
✓ "Turn the minutes below into a one-pager for executives. Separate
decided from open. Use only numbers that appear in the source."
```
More often than not this is **a question of thoroughness, not technique.** You
would have to say this much to a person too.
- The basics → [Getting the answer you wanted](/guide/ai-how-to-ask)
- In depth → [The four parts of a good instruction](/guide/ai-intent-context)
- Why instructions slip → [Why instructions do not simply stick](/guide/ai-steerability)
## 3. Discern — how to look at what came back
**The most frequently skipped move, and where most incidents happen.**
An AI answer is well written. Well-written text looks correct. So you read it and
move on.
There are three questions, and none substitutes for another.
```
"Does this source actually exist?" → invented? (hallucination)
"Same answer if I ask the reverse?" → taking my side? (sycophancy)
"Same answer if I swap attributes?" → leaning? (bias)
```
- What to watch for → [Things to watch out for](/guide/ai-cautions)
- Invention → [Why it is wrong so plausibly](/guide/ai-hallucination)
- Leaning → [Agreeing with you, and leaning](/guide/ai-sycophancy)
### Describe ↔ Discern is a loop
A good answer on the first try is rare. The way this actually works is
**discerning what is missing, then describing that back.**
```
Describe → Discern → "ah, I never said who reads this" → Describe → ...
```
**Skip discerning and the loop stops turning.** You then have a bad answer with
no idea why, and no amount of prompt fiddling improves it.
To run it yourself → [Rewriting instructions](/guide/ai-rewriting-prompts)
## 4. Take responsibility — what happens after it ships
The first three you do even alone. **The fourth appears when the work goes to
someone else.**
Three parts:
| | Question |
|---|---|
| **Disclose** | Does the recipient need to know this was made with AI? |
| **Protect** | Was the material that went in allowed to go in? |
| **Record** | If this becomes a problem later, what do I explain it with? |
None of the three are visible when you work alone. Which is why **lifting a
personal way of working straight into an organisation snags right here.**
- Disclose → [Saying what AI made](/guide/ax-disclosure)
- Protect → [Tools are permissions](/guide/mcp-security)
- Record → [Audit log](/guide/cn-audit)
## Where is your team weak?
```
□ People use AI, but what for is up to each of them → no 1. Delegate
□ A bad answer gets "well, that's AI for you" → no 2. Describe
□ Someone has pasted an answer through as is → no 3. Discern
□ Nobody knows who made what with AI → no 4. Responsibility
```
**Usually the third box is ticked first.** And an organisation missing the third
is missing the fourth too — if you never look, there is nothing to record.
These four moves are also the skeleton of this whole manual. The machine's side
of it continues in [The four properties of the
machine](/guide/ai-four-properties); the organisation's side continues in the
"Why transform" part.
---
## Check
**1. Which of the four moves do real incidents mostly come from?**
Answer
**3. Discern.** AI answers are well written, so they look correct and get read
past. However much you polish the prompt, skipping discernment leaves the
incident intact.
**2. Why does distinguishing automation, augmentation, and agency matter?**
Answer
**Because the same task gets a different number of human looks depending on the
mode.** Contract review as augmentation means a person looks every time; as
automation, nobody does. "Let's automate it with AI" is a sentence that skipped
this decision.
**3. What does "Describe ↔ Discern is a loop" mean?**
Answer
Good answers rarely arrive first try; you improve by **discerning what is missing
and then describing that back.** Skip discernment and you cannot tell why an
answer is bad, so changing the prompt changes nothing.
---
If the words are the obstacle, start here →
[Glossary](/guide/glossary)