What Most People Think the Skill Is
Ask most professionals what makes someone good at using AI, and they’ll say: prompting. Write better prompts. Use the right framework. Learn the magic words.
Prompting matters. But it’s roughly 20 percent of the equation. The other 80 percent is context — and most people never learn it.
Prompting vs Context Engineering
A prompt is what you ask. Context is everything the AI knows before you ask it.
Think about how you’d brief a new contractor. You wouldn’t hand them a task list on their first day with no background. You’d explain the project, the client, the standards you work to, the things that have gone wrong before, the outcome you’re trying to achieve. You’d give them context. Then the tasks make sense and the work is better.
AI works identically. A well-contextualised AI interaction produces dramatically better output than the same question asked without context — even when the prompt itself is mediocre.
What Context Engineering Actually Involves
Context engineering is the deliberate practice of preparing the information environment before your AI interaction begins. It covers five components:
1. Role and Expertise
Tell the AI who it is in this interaction — specifically. Not “you are a helpful assistant” but “you are a senior commercial lawyer in New Zealand advising a mid-size manufacturing company on their supplier contracts. You are direct, practical, and flag risk without over-lawyering.”
The more specific the role, the more the AI’s responses are shaped toward what you actually need.
2. Background and Situation
What does the AI need to know about your organisation, client, project, or situation? Provide it explicitly. The AI cannot infer what you haven’t told it, and it will fill gaps with generic assumptions — which are often wrong for your specific context.
3. The Real Objective
Not “write an email” but “write an email that persuades a cautious CFO to approve a $20K AI training budget before end of financial year — the CFO is risk-averse, focused on ROI, and has previously asked why we can’t just use free tools.”
The objective should include the outcome you need, the audience’s specific characteristics, and what a successful output looks like.
4. Constraints and Style
Word limits, tone, format, things to avoid, things to emphasise. The more specific your constraints, the more consistent and useful the output — and the less time you spend editing.
5. Examples of Good Output
If you have previous output that worked well, include it. AI models are excellent at pattern-matching to examples — often more effective than detailed instructions. Two good examples will frequently outperform two paragraphs of description.
The Context Document: How Serious Operators Do It
The most effective AI operators don’t rebuild context from scratch for every interaction. They build context documents — reusable files that contain everything the AI needs to know about a project, client, or domain.
Before starting any significant task, they load the relevant context document. The AI already knows the background, the constraints, the style requirements, and the objectives. The prompt itself can then be short, specific, and task-focused.
This is why two people using the same AI tool on the same task get radically different results. One rebuilds from scratch every time. The other has built up a library of context documents that makes every interaction more effective.
Why This Changes the Skill Equation
Context engineering shifts the advantage from people with the best prompts to people with the best preparation. And preparation is a skill that scales — every context document you build makes your next interaction better.
It also explains something that confuses a lot of people: why AI sometimes produces excellent output and sometimes produces garbage, even when the prompt looks similar. The difference is almost always in the context, not the prompt.
Where to Start
Pick one recurring task you do with AI regularly — a document type you produce, a kind of research you do, a communication format you write. Build a context document for it:
- Who you are and what role you’re playing in this task
- The purpose of the task and who the output is for
- What good output looks like — and what to avoid
- Two examples of output you’ve been happy with previously
- Any specific constraints (format, length, tone, things to never say)
Use that context document for your next five interactions on that task. The consistency of output will be noticeably different.
That’s context engineering. It’s not complicated — it’s just deliberate. And it’s the skill that separates people who find AI genuinely transformative from people who find it occasionally useful.
GenAI Training NZ covers context engineering as a core skill in all our workshops. If you want your team to move beyond prompting basics and build professional-grade AI workflows, get in touch.
Also see: Context Engineering — full overview | ChatGPT Training NZ | AI Training for Teams
