Learning objectives. Define prompt engineering accurately and without hype; identify the core components of a well-formed request (objective, task, context, constraints, output, success criteria); explain why prompt engineering is not about finding "the right words"; and write a first prompt using the HPC-10 Layer 1 elements deliberately.

Why this matters

Prompt engineering has a marketing problem: it's been sold as a collection of magic phrases, secret formulas, and one prompt that supposedly does everything. None of that survives contact with the research (Module 4 goes through it in detail). What actually works is much less mysterious and much more useful: say clearly what you want, give the AI what it needs to do it, and check the result. This module builds that habit from the ground up.

Plain-English explanation

Prompt engineering is the deliberate design of instructions and inputs that guide an AI system toward a desired outcome.

That's the whole definition. It is not about finding one magical sentence. It is closer to the everyday skill of briefing a new, capable colleague on a task: you wouldn't hand them one vague line and expect a great result, and you wouldn't bury them in forty pages of caveats either. You'd tell them what you're trying to achieve, what you need from them specifically, anything that matters for getting it right, and what "done well" looks like.

Core lesson

PLAIN ENGLISH — the components of a real request

Whether or not you write them all out every time, every effective prompt has these ideas present, even if only implicitly:

Objective — what you're actually trying to achieve (not just the activity — the outcome behind it)
Task — the concrete thing you're asking the AI to do right now
Context — what it needs to know about your specific situation
Constraints — rules, boundaries, things to avoid
Examples — what "good" looks like, shown rather than just described
Output requirements — format, length, structure, audience
Success criteria — how you (or the AI, if you ask it to self-check) will know it's actually right

Module 3 turns this into a structure you can apply directly. This module is about understanding why each piece matters, using HPC-10's Layer 1 (Intent: Purpose, Instructions, Constraints) as the anchor.

PRACTITIONER — objective vs. task, and why the distinction pays off

The most common gap in weak prompts isn't missing detail — it's missing the objective behind the task. "Write a marketing plan" is a task. "I need to convince a cautious co-founder that a £5,000 test budget for paid social is worth trying" is an objective. Give the AI only the task, and it produces a generic marketing plan template. Give it the objective, and it produces something actually persuasive, structured around the co-founder's likely objections — because now it knows why the document exists, not just what shape it should be.

This is the single highest-leverage habit in this whole module: before you describe the task, state the objective it serves. One sentence is usually enough.

ADVANCED — why wording matters less than structure

Module 4 covers the research in full, but the headline finding matters here too: Wharton Generative AI Labs' rigorous, repeated-trial testing found that surface-level wording changes (politeness, imperative phrasing) produced large swings on individual questions that mostly cancelled out across a full task set, while the model itself and the clarity of the underlying instruction mattered far more. The practical consequence: don't spend your effort hunting for the perfect phrasing. Spend it making the objective, task, and constraints genuinely unambiguous — that's the lever that actually moves outcomes.

Beginner example

Weak: "Write a marketing plan."

Better: "I run a five-person bakery in Leeds. I want a simple, realistic marketing plan for the next three months, focused on driving more weekday footfall (our weekends are already busy). I have no marketing budget beyond £200/month. Keep it to five concrete actions, not a strategy essay."

Notice what changed: an objective (weekday footfall, not vague "more marketing"), context (a five-person bakery, existing weekend strength), a constraint (£200/month, five actions not an essay), and an implicit success criterion (realistic, actionable).

Intermediate example

A marketing manager reuses this pattern for every campaign brief request: "Objective: [what business outcome]. Audience: [who]. Constraint: [budget/timeline/brand voice]. Format: [what I need back]." Four lines, every time, turns a five-minute back-and-forth into a first draft that's usable immediately.

Advanced example

A product team builds a shared prompt template for competitive analysis requests that explicitly separates objective ("inform a pricing decision" vs. "inform a positioning decision" produce different useful analyses from the same raw research) from task (gather and structure the information) — because they learned the hard way that skipping the objective produced technically accurate but strategically useless comparisons.

Business example

A solo consultant asked an AI repeatedly for "a proposal for [client]" and kept getting generic, forgettable documents. After restating the actual objective each time — "this proposal needs to overcome the client's specific hesitation about cost, which they raised on our call" — the same AI, with no other change, produced proposals that directly addressed the real obstacle instead of reciting a generic service list. Nothing about the AI improved. The objective moved from implicit (and invisible to the AI) to explicit.

Practical exercise

Take a prompt you've written recently that disappointed you. Identify which of the seven components (objective, task, context, constraints, examples, output requirements, success criteria) were actually present versus assumed. Rewrite it with the objective stated as its own explicit sentence, separate from the task. Run both versions and compare.

Common mistakes

Stating the task and assuming the objective is obvious (it usually isn't, to the AI).
Confusing a longer prompt with a better prompt — length is not the lever; clarity is.
Hunting for "magic words" instead of fixing an actually ambiguous instruction.
Treating a single good result as proof the prompt works (Module 11 is why this matters).

Expert insight

If you only take one habit from this module, take this: before you write the task, write the objective, as its own sentence. It costs ten seconds and it is the highest-return single change most people can make to how they prompt.

Knowledge check

Define prompt engineering in one sentence, without using the word "magic."
Distinguish an objective from a task, with an example.
Why did politeness-framing swings in the Wharton research "cancel out" across a task set, and what does that imply about chasing perfect phrasing?
List the seven components of a well-formed request.
Why is stating the objective explicitly often the highest-leverage single change to a prompt?

Module summary

Prompt engineering is the deliberate design of instructions and inputs — not a search for secret phrasing. Every effective request carries an objective, a task, context, constraints, examples, output requirements, and success criteria, even when some are implicit. The objective is the piece most often missing, and restating it explicitly is the cheapest high-value habit in this course.

Further exploration

Brown et al., Language Models are Few-Shot Learners (arXiv:2005.14165) — the foundational paper behind in-context learning, why AI systems can pick up a task from a well-specified request without retraining.

Prompt Engineering Fundamentals