Give AI a Job, Not a Question: Better Prompts Fast

Stop asking the AI a bare question. Hand it a job — a clear role plus a clear goal in one line — and watch generic answers turn sharp.

Most prompts start as a question. "What should I write in this email?" "How do I fix this paragraph?" "What's a good plan for this?"

Questions feel natural. But a bare question gives the AI no idea who it should be or what success looks like. So it answers like a polite generalist. Vague in, vague out.

Today's habit fixes that. You give the AI a job, not just a question. A job has two parts: a role (who the AI is) and a goal (what you want it to achieve). Put both in one line, and the quality jumps right away.

This is a technique called role prompting. It is one of the cheapest upgrades you can make. It costs a few words and pays back on every single reply.

Why a bare question gives you a bare answer

An AI model knows a little about almost everything. That sounds great. In practice, it is the problem.

When you ask a plain question, the model has no reason to pick one kind of answer over another. So it averages. It gives you the safe, middle-of-the-road response that offends no one and impresses no one.

Think about how this works with people. If you ask a stranger "How should I price this?", you get a shrug. If you ask "You run a pricing team at a software company — how should I price this?", you get real thinking. The role unlocks a specific kind of expertise.

A role does the same to the AI. It tells the model which slice of its knowledge to use. A "skeptical security reviewer" and a "friendly onboarding coach" answer the same question in completely different ways. The role chooses the lens.

The habit: role + goal in one line

Here is the whole move. Before your question, add one line that names the role and the goal.

The shape looks like this:

You are [a specific expert]. [Do this specific thing] for [this audience or purpose].

That is it. Two slots, one line. Then your details follow as normal.

Let me make each slot concrete.

Pick a role with real standards

A good role carries standards with it. "A marketing person" is weak. "A B2B email copywriter who writes for busy executives" is strong. The second one has clear ideas about length, tone, and what counts as good.

Aim for a role you could picture hiring. Add the traits that matter for your task: the audience, the focus, the way they work.

State a goal, not just a topic

A topic is "this email." A goal is "get this client to confirm the meeting time without sounding pushy." The goal tells the AI what winning looks like.

Always include who the output is for. "Explain this" and "explain this to a nervous first-time customer" lead to very different answers. The audience shapes the words.

Before and after

Watch what one line does. Same task, two prompts.

The first prompt gets you a generic template with five blanks and a robotic tone. The second gets you something you could nearly send as-is. Same effort to type. Very different result.

Here is another, from a different kind of work.

Notice the goal does extra work in the second one. "Three most important problems first" stops the AI from dumping forty tiny nitpicks. The role sets the standard. The goal sets the focus.

Match the role to the job

The best role is the one a real person would hire for that exact task. So change the role with the task. Don't reuse "helpful assistant" for everything.

Here are a few starting points you can adapt:

See the pattern? Each role carries a point of view. "Looks for risks first" or "cuts without losing meaning" tells the AI what to prioritize. That priority is doing real work for you.

You don't have to invent these from scratch every time. SurePrompts' AI prompt generator builds the role and goal into the prompt for you from a plain-English description, which is a handy way to see strong role lines and learn the pattern by example.

Don't overdo the role

A good thing has a limit. One strong role beats a crowded mix.

It is tempting to stack roles. "You are a lawyer, a poet, a data scientist, and a comedian." This usually backfires. The roles compete, and the output gets muddy. The AI tries to please all of them and serves none well.

Also resist roles that are too grand. "You are the world's greatest expert" adds nothing real. The model can't be more expert because you flattered it. Specific and honest beats big and vague every time.

Put it together

Let me show one full prompt using the habit, so you can copy the rhythm.

You are a careful financial coach who explains money to people
without a finance background. Goal: help me decide whether to pay
off a small debt now or keep the cash for emergencies.

Here are my numbers: [your details].

Give me a clear recommendation, the main trade-off in plain words,
and one risk I might be ignoring.

Read the first sentence on its own. Before the AI sees a single number, it already knows who to be and what success looks like. The rest of the prompt then has somewhere to land.

That front line is the whole habit. Role plus goal, then the details.

This pairs well with the habit from Part 2. A role tells the AI who to be. An example shows it what the finished work looks like. Together they cover most of what makes a prompt strong. In the next part, we go deeper on the power of a single example.

Try this today

Here is your tiny assignment. Take a prompt you already use — one you reach for often — and add a role-plus-goal line to the front. Nothing else.

Run the old version and the new version on the same task. Read them side by side. Keep the line that gave you the better answer.

You'll feel the difference fast. And once you do, writing that first line becomes automatic. That is the habit working. Want to check how much your upgraded prompt improved? Drop it into the free prompt scorer and watch the number move.

Keep going

Next → Part 4: Show, Don't Tell — One Example Beats Five Sentences

Or see the full Prompting Pro in 21 Days series.