Requirements Management: Attributes and Prioritisation

The syllabus identifies exactly two requirements management tasks AI supports: assignment of attributes, and prioritisation. Knowing that it is two, and which two, is itself examinable.

Assignment of attributes to requirements

Attributes document important metadata that lets stakeholders access relevant information throughout the project lifecycle. AI can streamline their assignment by analysing requirement content and proposing values based on established patterns and project-specific criteria.

Through textual analysis and comparison with similar requirements from current or previous projects, AI can suggest values for common attributes such as:

Priority
Complexity
Source
Responsible person

It can also identify missing attribute assignments and ensure completeness across large requirement sets, supporting the systematic documentation practices essential for effective requirements management.

Completeness checking is the quietly valuable half. Attributes are the part of requirements management that decays first: they are filled in diligently at the start of a project and increasingly forgotten as delivery pressure rises. A pass that reports which of four hundred requirements have no source recorded finds a real and common gap, and it needs no judgement to do so.

Proposing values is different in kind. A suggested priority derived from "similar requirements in previous projects" is a statement about other projects. Whether it holds here is a question about this project's stakeholders — and module 4's opening warning applies directly: where attributes are assigned with AI help, human judgement is required to confirm that critical factors such as business value and stakeholder approval status have been correctly evaluated. The assignments must accurately reflect project status and must not become misleading information.

That last risk is specific and worth naming. An attribute is metadata people rely on without re-deriving. A wrongly assigned "approved" status is worse than a blank one, because a blank field prompts a question and a wrong value stops one.

Requirements prioritisation

Generative AI can assist in prioritising requirements by analysing multiple criteria simultaneously and proposing rankings based on an evaluation of business value, urgency, effort, dependencies and other relevant factors.

Its distinctive strengths:

Processing a large quantity of requirements and stakeholder input to identify patterns and trade-offs that might be overlooked during manual prioritisation.
Ensuring systematic consideration of all prioritisation steps.
Generating comparative analyses among requirements, highlighting potential conflicts or dependencies that might influence prioritisation decisions.
Adapting prioritisation suggestions based on different stakeholder perspectives or changing project constraints.

The last point describes something genuinely hard to do by hand: re-ranking a large backlog from a different stakeholder's perspective, to see how much the ordering actually depends on whose view is taken. Doing that manually for three stakeholder groups across two hundred requirements is a week's work. It is also exactly the kind of comparison that makes a prioritisation negotiation productive, because it turns "we disagree" into "here is precisely where our orderings diverge."

Why prioritisation stays human

The syllabus is unambiguous. Effective prioritisation extends far beyond algorithmic analysis and requires deep understanding of stakeholder needs, business strategy and project context — elements AI cannot fully comprehend.

Three consequences follow:

Involving appropriate stakeholders in priority decisions remains essential. AI suggestions must be validated against their judgements and against strategic objectives.
Prioritisation must reflect genuine stakeholder needs rather than abstract metrics, to maintain alignment with the value-orientation principle (CPRE principle 1).
In iterative development, the dynamic nature of prioritisation demands ongoing human oversight to ensure AI recommendations remain relevant as project conditions evolve.

Module 4 introduced opacity with an example that lands squarely here: the AI proposes prioritising certain requirements but cannot explain the ranking logic, making review difficult.

Prioritisation is the RE activity where opacity does the most damage, and the reason is political rather than technical.

A priority order is a statement about whose needs come first. It is the output of a negotiation, and its legitimacy comes from the process that produced it — stakeholders were heard, trade-offs were made explicit, someone with authority decided. When the order is questioned months later, the answer is that reasoning.

An AI-produced ranking has an output but no such process. If it cannot explain why requirement 47 outranks requirement 12, then there is nothing to defend when the sponsor of requirement 12 objects. The Requirements Engineer is left either overriding the tool arbitrarily or defending a conclusion they cannot justify — and both damage the trust the trust-builder role in module 5 exists to protect.

This is why the syllabus routes prioritisation so firmly back to stakeholder involvement. The useful contribution is the comparative analysis — the dependencies, the conflicts, the trade-offs surfaced across a large set. That is input to a negotiation. The ranking itself has to come out of the negotiation, because that is where its authority comes from.

❓ Knowledge Check

Which two requirements management tasks does the syllabus identify as supported by AI?

📚 Flashcards1 / 5

Term

Attribute assignment support

Click to flip

Definition

Analysing requirement content to propose values for attributes such as priority, complexity, source or responsible person, by comparison with similar requirements from current or previous projects, and identifying missing assignments across large requirement sets.

Click to flip back

💡Key Takeaway

AI supports exactly two requirements management tasks. For attributes it proposes values and, more reliably, finds the assignments that are missing — but a wrong attribute is worse than a blank one, because it stops the question a blank field would prompt. For prioritisation it processes volume, surfaces dependencies and re-ranks from different stakeholder perspectives, which is valuable input to a negotiation. The ranking itself must come out of that negotiation, because that is where its authority lives.