When answers become cheap, knowing what you are trying to achieve becomes extraordinarily valuable.
There is a lot of talk about the skills students will need in an AI world.
Coding. Prompt engineering. Critical thinking. Data literacy. Creativity.
All useful.
But I think the most important skill may be simpler — and harder.
Intent.
Or, to use a consulting term: the brief.
Know what you are trying to achieve, and learn how to articulate it, and you will do well in the AI future.
That sounds obvious. It isn't.
AI can do the work. It cannot decide what is worth doing.
AI is becoming remarkably good at execution. It can produce a report, itinerary, lesson plan, presentation, image, computer program or marketing campaign in minutes.
But faster execution makes the quality of the original intention more important, not less.
If you ask for a report, you will get a report. That does not mean it is the report you needed. If you ask for a holiday, you will get an itinerary. That does not mean your family will enjoy it. If you ask for a business idea, you will get several. That does not mean any of them should exist.
The result can be polished enough that we forget to ask whether it solves the right problem.
The travel agent problem
Consider the travel agent.
Traditionally, travel agents took input from punters who knew little about airlines, fare rules or the travel industry — and who were fortunately not always aware of the commissions and incentives sitting behind the recommendations.
Self-service took a large slice of that business. Or, put another way, the industry replaced many of the people in Flight Centre with digital marketing spend and booking websites.
Most of us can now book a flight to Melbourne without help.
But creating a brief for a family holiday is a different exercise.
Booking is the easy part. The real work is deciding what a good holiday means for this family.
Think about your own family. One person wants rest. Another wants adventure. A teenager needs Wi-Fi and would rather not admit that this matters. Someone has a bad knee. Someone hates early starts. There is a budget, a school calendar, dietary needs, passport expiry dates and a limited tolerance for changing hotels.
Then come the things that are harder to specify. Should the holiday stretch the children a little? Expose them to another culture? Create time together? Include something none of you would have chosen individually?
Booking is the easy part.
The hard part is deciding what a good holiday means for this particular group of people, at this particular point in their lives.
That is intent.
We already pay people to help us form intent
The travel agent is not an isolated example. Many professional services are partly exercises in helping a client discover and articulate what they actually need.
The architect
A client might begin with: “We need another bedroom.” A good architect asks what has changed. Is the real need privacy for a teenager, space for an ageing parent, somewhere to work, or a house that can adapt over the next decade?
The first request names an output. The brief identifies the life the building must support.
The doctor
A patient may arrive wanting a prescription, a scan or simply for the pain to stop. The clinician has to uncover symptoms, history, risk, priorities and trade-offs. The first requested intervention may not be the right one.
The skill is not merely answering the question. It is working out what question should be answered.
The film director
A film set contains people who are far better than the director at lighting, sound, costume, editing and camera operation. The director's value is not that they personally perform every task. It is that they hold and communicate the intent of the whole.
The manager
A weak manager assigns activity: “Prepare twenty slides on the market.” A better manager explains the decision those slides must support, the audience's concerns, the evidence required and what remains unknown.
The second version gives the person doing the work permission to think.
That distinction matters even more when the person doing the work is partly an AI.
Intent is not the same as a clever prompt
Prompt engineering is useful. But I suspect we overestimate the importance of knowing the right words to type and underestimate the importance of knowing what we mean.
Intent comes before prompting.
A good brief usually contains at least seven things:
1. Outcome
What should be different when this work is finished?
2. Audience
Who is it for, and what do they already know, believe or need?
3. Context
What situation produced the request?
4. Constraints
What limits must be respected: time, money, evidence, ethics, accessibility or format?
5. Trade-offs
If we cannot maximise everything, what matters most?
6. Success
How will we judge whether the result is useful?
7. Responsibility
Which decisions may the AI support, and which must remain human?
For example, “Write me an essay about climate change” barely expresses intent.
“Help me compare two explanations for why Australian cities struggle to reduce transport emissions. I need to identify the assumptions behind each explanation, use credible evidence, and finish with the questions I would investigate next” is different. It specifies a purpose and a standard of thought. It also leaves the learning with the student.
What intent looks like when you are working with AI
Intent is not a magic phrase that you insert into one perfect prompt. It is a shared understanding that develops through conversation.
Tell the AI what you are trying to achieve, why it matters, what must remain yours, and how you will judge the result. Then invite it to ask questions, expose assumptions and offer alternatives.
Here are three examples.
1. A student assignment
The output request is easy:
The intent-led version is different:
“I have to write a 1,500-word university assignment examining whether social media harms democracy. My learning goal is to compare competing causal arguments and judge the strength of their evidence. Do not write the assignment for me. Start by asking three questions about my tentative view. Then help me map two credible positions, identify the evidence I should look for, and challenge my assumptions. Flag every claim I will need to verify. Finish with a possible structure that I can turn into an argument in my own voice.”
This does more than improve the eventual essay. It protects the purpose of the assignment: the student must still form a view, evaluate evidence and write.
2. Writing a song
Again, the output request is simple:
An intent-led creative conversation might begin like this:
“I am developing an acoustic song about the excitement and guilt of leaving a regional town for the city. It is for young adults facing the same decision. I want it to feel bittersweet rather than triumphant, told in the first person, using concrete images from a railway platform. I want to remain the songwriter. Help me explore three possible emotional angles, ask which one feels most truthful, and suggest some chorus structures and chord moods. Do not imitate a particular artist or produce a finished lyric unless I ask for one.”
The AI now knows the emotional destination, audience, perspective and boundaries. It can expand the creative space without quietly taking authorship of the work.
3. Researching a hobby
Suppose the initial request is:
Useful, perhaps. But it is likely to produce a generic collection of facts.
“I want to start observing Australian magpies around Sydney with my twelve-year-old. My real goal is to spend time outdoors together and learn how to notice animal behaviour—not simply collect facts from the internet. We want to understand their different calls, social groups, territories and whether individual birds can be recognised over time. Help us design a three-week weekend observation project with simple field-notebook prompts. We will observe respectfully without feeding, approaching nests or interfering with the birds. Ask any questions about our location or season that would materially change the plan, include sensible advice for swooping season, and point us towards credible Australian sources. Clearly separate established research from interesting possibilities we would need to investigate further.”
The subject is still magpies, but the intent is connection, careful observation and learning how knowledge is built. That produces a very different answer.
In each case, intent changes the role of AI. It stops being a vending machine for finished outputs and becomes a partner in clarification, exploration and challenge.
The academic case for intent
There is a substantial research trail behind this idea, even if researchers use different language.
Goal-setting research has repeatedly found that specific, challenging goals tend to produce better performance than vague exhortations to “do your best,” provided people have the capability, commitment and feedback needed for the task (Locke & Latham, 2002). A brief turns a general aspiration into something against which choices can be made.
Design researchers Kees Dorst and Nigel Cross showed that creative work is not a neat march from a fixed problem to a solution. Designers develop their understanding of the problem and the possible solution together (Dorst & Cross, 2001). That is an important qualification: knowing your intent does not mean freezing the brief at the beginning. A good brief develops as you learn.
Research with university students solving ill-structured problems has also found that prompts which help learners integrate knowledge and attend to the problem-solving process can improve how they represent the problem and develop solutions (Chen, 2010). In other words, the questions we ask before and during execution affect the quality of the thinking.
Early research on teaching prompt engineering to undergraduates suggests that even a short, structured intervention can improve AI knowledge, confidence and prompting ability (Woo et al., 2024). This is promising — but the deeper educational opportunity is not to teach students a bag of prompt tricks. It is to teach them how to clarify purpose, surface assumptions, specify constraints, inspect results and revise their direction.
The OECD Learning Compass 2030 uses a useful metaphor: students need to navigate unfamiliar contexts and find direction, rather than simply receive fixed instructions. AI makes that distinction urgent. When answers are cheap, choosing the direction becomes valuable.
How do students learn intent?
Probably not by listening to a lecture about it.
They learn it by being given incomplete, messy situations and having to form the brief before producing the work.
Ask students to interview a real “client,” even if that client is another student. Require them to distinguish a requested output from the underlying need. Ask them to name stakeholders, constraints and trade-offs. Let them use AI — but assess the quality of the brief, the decisions they made and the way they evaluated the result.
Most importantly, ask them to show how their intent changed.
- What did you think the task was at the start?
- What did you learn that changed your view?
- Which constraint mattered more than you expected?
- What did the AI produce that was technically competent but wrong for your purpose?
- What decision did you keep for yourself, and why?
That is a much richer demonstration of learning than submitting a polished output whose origins nobody can see.
The brief is a human act
The future will not belong simply to the people who can make AI produce the most material.
It will belong to people who can decide what deserves to be produced; who can articulate outcomes without pretending away uncertainty; who can represent the needs of other people; who can make trade-offs visible; and who can recognise when a beautifully executed answer is answering the wrong question.
AI can help us explore. It can challenge a brief, expose missing constraints, propose alternatives and accelerate the work.
But someone still has to choose the destination.
That is the skill.
Intent.
Or, if you prefer the consulting term: the brief.
References
- Chen, C.-H. (2010). Promoting college students' knowledge acquisition and ill-structured problem solving: Web-based integration and procedure prompts. Computers & Education, 55(1), 292–303.
- Dorst, K., & Cross, N. (2001). Creativity in the design process: Co-evolution of problem–solution. Design Studies, 22(5), 425–437.
- Locke, E. A., & Latham, G. P. (2002). Building a practically useful theory of goal setting and task motivation: A 35-year odyssey. American Psychologist, 57(9), 705–717.
- OECD. The OECD Learning Compass 2030.
- Woo, D. J., Wang, D., Yung, T., & Guo, K. (2024). Effects of a prompt engineering intervention on undergraduate students' AI self-efficacy, AI knowledge and prompt engineering ability: A mixed methods study. Preprint.