Prompting Patterns

Output quality depends not only on context — what the model should know — but also on how instructions are phrased. That is the territory of educational objective 3.2.

Prompting patterns are reusable structures that help formulate prompts consistently and effectively. That sentence is the general purpose the exam asks you to name. Patterns prescribe a structure.

Role-Task-Format

The most common pattern, and the one to know cold, is RTF:

Role — a description of the role the AI should assume.
Task — the specific job to perform.
Format — the expected structure of the output.

Applied to a real RE job:

Role:   You are a Requirements Engineer reviewing a supplier's
        specification for a hospital patient-monitoring system.

Task:   Read the requirements below and identify every requirement
        that is not testable. For each one, say why it cannot be
        tested and propose a measurable alternative.

Format: A table with three columns — original requirement, reason it
        is untestable, proposed measurable version. No preamble.

By combining these three elements, you guide the model away from generic answers and towards precise, actionable results. The Format element in particular does more work than people expect: specifying "a table with three columns, no preamble" removes an entire round of reformatting.

Other patterns worth recognising

Many patterns exist. The syllabus names two more as important examples:

Chain-of-Thought. Ask the AI to explain its reasoning step by step before giving the final result. In RE this supports transparency and traceability — for example when analysing conflicting requirements or evaluating design alternatives, where the intermediate steps are the part a stakeholder needs to see.

Keep module 2 in mind here. Chain-of-thought works because the model has seen many step-by-step explanations in its training data and can reproduce that format. It improves results on tasks that benefit from structure. It does not turn prediction into reasoning.

CRISPE. A more detailed pattern: Context, Role, Instructions, Steps, Parameters, Examples. Use it when a task is complex enough that RTF's three slots leave too much unsaid.

🔗 Match the Pairs

Role-Task-FormatDrop here

Chain-of-ThoughtDrop here

CRISPEDrop here

The three classes of benefit

The syllabus groups the advantages of prompting patterns into three categories. This structure is very examinable, because it is one of the few explicit taxonomies in the unit.

Practical benefits. Patterns increase the usability of answers by providing clear instructions upfront. Defining the desired format reduces the need for post-processing and saves time. Well-structured prompts shorten the number of iterations required to reach a useful result.

Cognitive benefits. Using a pattern forces you to think carefully about your needs before asking. By making role, task and format explicit, ambiguity is minimised and the reasoning behind a prompt becomes transparent and reproducible.

This benefit is easy to underrate and is arguably the most valuable one for a Requirements Engineer. It is the same discipline that makes a well-designed interview question better than an open-ended chat: the structure forces the questioner to decide what they are actually asking.

Methodological benefits. Patterns can be reused across different tasks, ensuring consistency and lowering the learning curve for new users. Within a team they create a shared standard for working with AI. Combined with prompting techniques, they provide both a stable structure and situational flexibility, leading to more reliable and adaptable results.

The methodological benefit becomes concrete when a pattern stops being something one person types and becomes something a team stores.

A requirements team that agrees on a single RTF prompt for converting stakeholder notes into user stories gets several things at once:

Every story that reaches the backlog has passed through the same instruction, so format drift between team members disappears.
When the output turns out to be systematically wrong in some way — say it keeps inventing acceptance criteria that nobody asked for — there is one artefact to fix, not a habit spread across five people's chat histories.
A new team member becomes productive by using the team's prompt rather than by developing their own prompting instincts.
The prompt becomes reviewable. It can be version-controlled, discussed in a retrospective, and improved deliberately.

That last point is the reason the syllabus calls this a methodological benefit rather than a practical one. A stored, shared, improvable prompt is a piece of method, and it behaves like any other piece of method the team owns.

❓ Knowledge Check

Which of these is the syllabus's stated general purpose of prompting patterns?

📚 Flashcards1 / 7

Term

Prompting pattern

Click to flip

Definition

A reusable structure that helps formulate prompts consistently and effectively. Patterns prescribe structure, in contrast to techniques, which influence interpretation.

Click to flip back

💡Key Takeaway

Prompting patterns are reusable structures that make prompts consistent and effective. Role-Task-Format is the canonical example; Chain-of-Thought and CRISPE are two more. Their benefits fall into three classes the syllabus names explicitly: practical (fewer iterations, less post-processing), cognitive (forces clarity before asking), and methodological (reusable, consistent, a shared team standard).