Your first try

What you will learn The reading stops here. This chapter is for doing. Follow the three below exactly. It takes five minutes.

Before you start

Nothing to prepare. A browser is enough.

If you use HyperTeams Connect, entering a workspace drops you straight into a chat box (getting started). If you use a different AI service, everything below applies just the same.

It is fine to get it wrong. Nothing breaks because you phrased something badly. If you do not like the answer, say so again. Go in relaxed.

First — just talk to it

Type this into the chat box. Copy and paste is fine.

text

Hello. This is my first time using AI.My job is (write your role here, e.g. office administrator at a small company).Give me 5 things in my work that AI could help with.

What to notice

The answer comes back fitted to your role. Write "office administrator" and you get office administration.
Search would have given you a generic "how to use AI" article.

This is the "answer fitted to your situation" from what AI is.

Second — ask a follow-up

Once the first answer arrives, do not start over. Continue with this:

text

Tell me more about the third one.In the order I could actually try it today.

What to notice

It remembers the earlier conversation. "The third one" is enough.
This is the big difference from search, where you retype everything each time.

The thing to internalise here A conversation with AI is not a single shot. If you do not like the answer, keep going: "shorter," "with an example," "as a table." It is a conversation.

Third — give it a real piece of work

Now something real. Pick the one closest to your situation below.

If you have an email to write

text

Draft an email for the situation below.Polite but short — about five sentences.To: (e.g. a supplier contact)Situation: (e.g. I sent a quote last week and haven't had a reply.           I'd like to follow up politely.)

If you have something long to read

text

Summarise the following in 3 lines.Then separately tell me anything I absolutely need to know.(paste what you want summarised here)

If you have something to organise

text

Turn the following into a table.Columns: (e.g. Item / Owner / Deadline).(paste what you want organised here)

When you do not like the result

This is where most people give up. It is also the most important step.

If you do not like it, say so. Like this:

Complaint	Say this
Too long	"Cut it to half the length"
Too stiff	"Rewrite it in a warmer tone"
Does not fit my situation	"We're (situation), so that won't work. Suggest again"
I don't understand it	"Explain it so a ten-year-old would get it"
Too generic	"Our company is (specifics). Redo it for that"

How it actually goes

text

You: Tidy up these meeting notes (+ paste notes)AI:  (tidies them up)You: Pull out just the decisionsAI:  (extracts the decisions)You: Group them by ownerAI:  (regroups by owner)

Three turns and it is what you wanted. Do not strain to write the perfect request up front. Narrowing as you go is faster.

Once you have done it

If you got this far, you have the basics. In summary:

Write like you are talking to a person (do not compress into keywords)
Give it the situation
If you do not like it, keep talking

That is all of it. The rest is technique, and that is the next chapter.

Check yourself

1. What should you do when you do not like AI's answer?

Answer

Keep talking. Rather than starting a new conversation or giving up, correct it — "shorter," "that doesn't work because we're in this situation." Because it is a conversation, it fixes things while remembering what came before.

2. Why does "tell me more about the third one" work?

Answer

Because it remembers the earlier conversation. Search makes you retype from scratch every time; AI continues knowing everything said so far.

3. Why is there no need to write a perfect request up front?

Answer

Because narrowing as you go is faster. As in the meeting-notes example, three quick exchanges get you to the shape you wanted sooner than agonising over one perfect first message.

Next, five techniques that make the same request produce much better results → Getting the answer you wanted