Clarify Your Intention Before Writing Your Prompts
You have seen what AI can do, and where its limits are. Now let’s turn to the skill that changes everything: how to talk to it so it actually helps you. Good news, this is not a magic trick, and it is not a hunt for the one perfect prompt. It is a framing skill, and you can build the reflex in a single chapter.
This is the whole point of prompt engineering: turning a vague request into a precise instruction. Let’s build that habit now.
Define Your Objective, Audience, and Format
Before asking for anything, run a quick test in your head. Imagine you type: “Suggest something nice for my vacation.” What happens? The AI most likely fires back a list of questions, because the request is too vague to act on. You could answer those questions one by one, but that is a slow ping-pong. Better to get it right the first time.
Answer three questions before you write a single word:
Objective: what do I want to obtain?
Audience: who am I writing this for?
Format: what shape should the result take?
Answer all three inside your request and the reply changes completely. You save time, and you avoid generic answers. Here is a concrete comparison, using your Operation Great Escape project. Suppose you want a plan for the week.
Without framing you ask: “Plan my week away with friends.” You get a generic, one-size-fits-all itinerary, too long or too short, that ignores your real constraints.
With Objective-Audience-Format, you would ask for: a decision-ready 7-day itinerary in France (Objective), for a group of four friends with mixed budgets and no travel expertise (Audience), presented as a table with one row per day and columns for the main activity and an estimated budget (Format). Now you get something you can actually use, or adjust in seconds.
Write Requests the AI Can Actually Use
Once you have your objective, audience, and format built into your prompt, tighten the wording. Remember: what you do not write, the AI cannot guess. To be genuinely usable, a request should carry four elements.
A clear action verb: generate, summarize, analyze, rephrase.
Enough context: the information the AI needs, without drowning it.
Explicit constraints: length, tone, structure.
A success criterion: how you will judge that it is good.
Watch the difference on one request. The vague version reads: “Write a good summary for Operation Great Escape.” There is nothing there to work from, so the AI fills the gaps itself and hands you something generic.
This version names all four elements: “Write a short summary to present Operation Great Escape to the group. About 120 words, friendly and motivating tone, ending on a clear call to confirm dates. The goal is to make three hesitant friends want to reply today.” Same task, far better result.
Break Complex Tasks into Sub-Tasks
As you get comfortable, you will be tempted to hand the AI a whole mission at once: “Compare three accommodations and build a budget and plan a weather-proof itinerary.” Tempting, but it rarely works as well as you hope. The AI performs better on one focused task than on a request that stacks several goals, exactly like a person would.
So slice the big task into a short sequence. Take that same overloaded request and break it up:
First, ask the AI to help you set your criteria: what matters for the accommodation (price, location, group size)? Let it help you frame the comparison.
Then, ask it to compare the three options against those criteria, and only that.
Finally, ask it to draft a day-by-day itinerary with a rainy-day alternative for each day.
Write one request per step, not three at once. Each answer feeds the next, and you assemble and check the final result yourself.
Your Turn!
You are moving Operation Great Escape forward. You need to send the group one message today to lock three decisions: the destination, the budget per person, and the type of accommodation. The message must be short and make people want to reply.
Instructions
Step 1 - Think before you write
In two lines, write the deliverable you want (ex: the group message).
Note two non-negotiable constraints (tone, length, format, or audience).
Step 2 - Build your prompt
Objective: what concrete result do you expect?
Audience: who will read this message?
Format: what structure, for example three short bullets, one per decision?
Run the prompt with all three parts included.
Step 3 - Iterate and derive a personal rule
Make one single change (adjust the audience, the format, or one constraint) and run it again.
Note what improved the result the most, and write it down as a rule you will reuse.
When you are done, check your answer against the model solution found at the end of this chapter.
Let’s Recap!
Defining the Objective, Audience, and Format of the result you want forces you to clarify your intention before you delegate and yields a better quality result.
A usable request combines an action verb, enough context, explicit constraints, and a success criterion.
When a request piles up several “and”, break it into a sequence of shorter prompts.
Prompt chaining is more reliable and easier to correct than one overloaded request.
Change one parameter at a time so you can see what actually improved the result.
You now have a clear frame for getting precise answers. Next, we decide what is worth delegating, and with how much risk.
Model Solution
The goal is not to produce the one “right” prompt, but to see how defining the Objective, Audience, and Format changes the result. Here is a worked path.
Step 1 - Example framing. Deliverable: one group message proposing the destination, a budget per person, and an accommodation type. Constraints: under 120 words, warm and casual tone.
Step 2 - Example Objective-Audience-Format prompt. “Write a group message to my four friends (Audience) proposing our destination, a budget of around 400 euros per person, and a shared rental (Objective), as three short bullets ending with a question that invites a quick yes or no (Format).” This works because every dimension is filled in. A prompt with no Format might have produced a long paragraph nobody reads to the end.
Step 3 - Example iteration. If you change the Audience from “close friends” to “friends plus two people I barely know”, the AI will likely soften the tone and add a short line of context for the newcomers. Changing a single element visibly changed the message.
