What gets better
What you will learn The answer to "fine, it's good, but what do I use it for." Five real tasks, with the actual input and the actual output.
The big picture first
The work AI is good at has something in common.
Making 1 out of 0, and making many into few.
People spend most of their time on writing the first line in front of a blank screen and reading something long all the way through. Those two are exactly where AI is strongest.
Conversely, polishing a 1 into a 10 and judging whether something is right are still better done by people. So the good pattern is AI drafts, a person judges.
Now the five, one at a time.
1. Long into short — cutting reading time
The situation
A supplier sent back a revised 12-page contract. You need to know what changed and you have no time to read it.
What you write
What comes back
What changed
Instead of reading twelve pages you learn where to look and check only those parts. The judgement is still yours — whether what AI flagged is actually a problem is for you to decide.
It did not cut your time. It decided where your time should go.
2. Writing the first line — killing the blank screen
The situation
A delivery slipped by a week. You need to write an apology to the customer and the first sentence will not come.
What you write
What comes back
What changed
This draft does not go out as-is. You adjust it to your company's voice, put in the right name, add anything missing. But you did not start from a blank screen.
Editing is far faster than writing.
3. Scattered into ordered — making a table
The situation
You took notes in a meeting but they are all over the place. You have to tidy them up and share them.
What you write
What comes back
What changed
Same information, now in a form you can read. This job is not hard, but it takes time. Simple work that takes time is exactly the work worth handing off.
4. Another language — situational, not dictionary
The situation
You need to email an overseas supplier in English. Machine translation comes out subtly wrong.
What you write
What comes back
Feed the original straight into a translator and "we'd like to receive a
catalogue" comes out as We want to receive a catalogue. The meaning is right,
but it is not how a first approach to a supplier reads. The version above is
rewritten into the shapes English business email actually uses —
Could you also advise….
What changed
A translator moves sentences across; AI writes for the situation. It honours conditions like "formal" and "first contact." Rather than transposing your original sentence by sentence, it rewrites it the way it would be written natively.
5. When you are stuck — generating a list
The situation
You have been told to "think about how we get more new customers." You have no idea where to start.
What you write
What comes back
Ten of them. The first three look like this.
What changed
Eight of the ten may be things you already knew or that do not fit your situation. But if two are usable, that is a win. This is for generating a starting point for thinking.
Choosing from ten is easier than squeezing ideas out of a blank page alone.
What the five have in common
graph TD
A["Where people spend time"] --> B["Reading something long"]
A --> C["Writing the first line"]
A --> D["Changing the shape of things"]
B --> E["AI makes a draft"]
C --> E
D --> E
E --> F["A person judges and polishes"]Every one is AI drafts, a person judges. AI is not producing the final deliverable.
The most common mistake Using what AI gave you as-is. A draft is a draft. Why, in things to watch out for.
Common questions
"My job isn't like any of these."
Those five are only examples. The test is this:
If words move through it, it is a candidate.
Count how much of your day is reading, writing, organising, or explaining documents. In most desk jobs it is more than half.
"I tried it and the results were mediocre."
Usually because the question was too short. Look at the examples again — every one describes the situation and attaches conditions. The technique is in getting the answer you wanted.
Check yourself
1. What do the tasks AI is good at have in common?
Answer
Making 1 out of 0 (first lines, generating lists) and making many into few (summarising, organising). Judging correctness and putting on the final polish are still better done by people.
2. In the contract example, what was actually reduced?
Answer
Not reading time — the time spent finding what to read. The judgement is still yours; whether the three clauses it flagged are really problems is something you have to check.
3. How do you tell whether your work is a candidate for AI?
Answer
Ask whether words move through it. If you read, write, organise, or explain, it is a candidate.
Now it is your turn. The next chapter is a first conversation you can simply follow along with → Your first try