1. Normalize the sources
Outcome: Every requirement has provenance.
Tasks
- →Collect the documents
- →Assign source IDs
- →Extract atomic requirements
Checks
- ✓No requirement exists without a source or an inference label
Core•8–14 hours
AI-assisted Requirements Audit
Audit requirements with traceability, ambiguity detection, gap and risk analysis, and verified acceptance criteria.
requirements analysistraceabilitygap analysisrisk framing
Scenario
Task
A team has a conflicting set of requirements spread across several documents. AI helps structure the material, but every conclusion must be linked to a source or explicitly marked as an assumption.
Step-by-step execution
2. Find gaps and conflicts
Outcome: Problems are classified and prioritized.
Tasks
- →Detect duplicates
- →Find contradictions
- →Flag missing controls
Checks
- ✓Every finding has evidence
- ✓Severity has a rationale
3. Build acceptance criteria
Outcome: The criteria are testable.
Tasks
- →Rewrite vague statements
- →Add measurable outcomes
- →Link risks
Checks
- ✓No uncontrolled terms such as fast or convenient remain
4. Run stakeholder review
Outcome: Uncertainty is converted into a decision or an open question.
Tasks
- →Prepare a decision log
- →Record unresolved items
Checks
- ✓AI inference is never presented as fact
Acceptance criteria
- 100% of requirements have provenance
- Conflicts have severity
- Acceptance criteria are testable
- Open questions are separated from conclusions
Assessment rubric
How the result is assessed
Passing score: 70/100 · Distinction: 90/100
Provenance and traceability
Every requirement has a source or an inference label.
Insufficient
Conclusions are not linked to sources.
Competent
Requirements have IDs and sources.
Strong
Versioning, lineage, and change impact are documented.
Evidence required
- ✓ Link to an artifact or code
- ✓ README with decisions and trade-offs
- ✓ Evidence of completed checks
- ✓ Traceability matrix
Gap/conflict analysis
Duplicates, contradictions, and missing controls are classified.
Insufficient
Findings are subjective or lack evidence.
Competent
Findings have severity and rationale.
Strong
Deduplication, impact mapping, and stakeholder validation are present.
Evidence required
- ✓ Link to an artifact or code
- ✓ README with decisions and trade-offs
- ✓ Evidence of completed checks
- ✓ Gap and conflict register
Acceptance criteria
Criteria are measurable and testable.
Insufficient
Criteria are vague.
Competent
Core outcomes are measurable.
Strong
Criteria are linked to risks, tests, and ownership.
Evidence required
- ✓ Link to an artifact or code
- ✓ README with decisions and trade-offs
- ✓ Evidence of completed checks
- ✓ Examples of rewritten criteria
Decision readiness
Open questions, assumptions, and decisions are separated.
Insufficient
Assumptions are presented as facts.
Competent
Statuses are explicitly separated.
Strong
Owner, deadline, and resolution path are defined.
Evidence required
- ✓ Link to an artifact or code
- ✓ README with decisions and trade-offs
- ✓ Evidence of completed checks
- ✓ Decision log