Skip to content

Policy Assessments

Where to find it: Policy Governance, then Existing Policy Assessments (/governance/assessments).

The Policy Assessment module (also known as Upload & Assess) allows organizations to evaluate existing internal policy documents against certified target expectations (such as Content 2.0 governance packs or template versions) to determine structural policy coverage.

Requires the policy_governance entitlement.


The Assessment Pipeline

The assessment engine executes a deterministic, multi-stage evaluation pipeline:

  1. Source Document: An existing internal policy version or an uploaded external document (.pdf, .docx, .txt). For uploads, the binary is immutably retained as a DocumentAsset pinned to its SHA-256 hash.
  2. Structural Extraction: The extraction worker decomposes the source text into ordered candidate segments (clauses, paragraphs, and numbered provisions).
  3. Target Expectation Snapshot: The assessment freezes an immutable snapshot of atomic requirements from a released target template version or pack version. Mutable drafts can never be targets.
  4. Mapping: Segments are aligned to target expectations. Mapping proposals may be generated manually or via advisory AI assistance.
  5. Human Review: An authorized reviewer (policy_assessment.review) evaluates mappings, accepting, rejecting, or modifying alignments.
  6. Result Determination: The reviewer records the authoritative coverage result for each expectation.
  7. Recommendations: Deficiencies generate actionable gap recommendations that route to the Gap Workbench.

Assessment Lifecycle States

Every assessment moves through four formal lifecycle states:

StatusMeaning
draftAssessment created, source document linked or uploaded, extraction pending
extractedStructural extraction complete; candidate segments and target expectations generated
in_reviewReviewer actively adjudicating mappings and setting coverage results
completedAuthoritative assessment finalized and sealed; permanently immutable

Completed assessments are immutable

Once transitioned to completed, an assessment is locked (ASSESSMENT_COMPLETED). Its segments, mappings, results, and summary statistics can never be modified. To evaluate revised policy text or new standards, users must create a new assessment record, preserving a tamper-proof historical audit trail.


Authoritative Coverage Result States

For each individual target expectation, the reviewer assigns one of five authoritative result states:

Result StateMeaning
addressedThe source policy contains clear, explicit provisions that fully satisfy the target expectation
partially_addressedThe source policy references or addresses the intent, but omits key operational criteria or safeguards
missingThe source policy contains no provisions corresponding to the target expectation
not_applicableThe expectation does not apply to this organization or operating scope (requires mandatory justification)
unassessedDefault initial state; pending reviewer evaluation

Core Evaluation Invariants

  • A mapping does NOT imply "addressed": The existence of a mapping between a clause and an expectation merely records semantic proximity. It does not certify that the clause satisfies the expectation.
  • Not Applicable requires rationale: An expectation cannot be marked not_applicable without recording a documented business or jurisdictional justification.
  • AI suggestions are strictly advisory: Machine-learning recommendations are clearly labeled as advisory. Only an authorized human holding policy_assessment.review can establish an authoritative result.

Candidate Mapping States and Reviewer Decisions

Each expectation-to-segment mapping record carries a canonical stored mapping_state:

Stored mapping_stateDescription
manualMapping created manually by a human compliance officer
ai_suggestedMapping proposed by the advisory AI extraction worker
acceptedCandidate mapping reviewed and confirmed by an authorized human reviewer
rejectedCandidate mapping discarded by reviewer as non-relevant

During adjudication, reviewers submit explicit API decisions that transition mapping records:

Reviewer Decision (API)Resulting Stored State / Action
acceptTransitions mapping_state to accepted
rejectTransitions mapping_state to rejected
modifyUpdates mapped segment_id or reviewer rationale without changing state
reopenResets mapping_state to ai_suggested (if AI origin) or manual

Gap Recommendations

When an expectation is determined to be missing or partially_addressed, the assessment engine generates structured recommendations:

Recommendation TypeRecommended Remediation Action
amend_existing_statementUpdate existing policy clause to incorporate missing criteria
add_statementInsert a new atomic statement into the current policy section
add_standard / add_procedureDraft supporting standard operating procedures
create_new_policyCreate a standalone corporate policy to address systemic deficiency
mark_naFormally record a justified out-of-scope determination
raise_exceptionRequest a governed, time-bound policy exception
keep_map_existingConfirm existing text is sufficient without editorial modification

These recommendations flow directly into the Gap Workbench for assignment and tracking.


Semantic Doctrine

Semantic Doctrine: Policy Coverage != Compliance

OrviQ enforces strict separation between structural policy coverage and operational compliance:

  • Terminology is Policy Coverage: The metric produced by this module is strictly Policy Coverage %, never "Compliance %", "Assurance %", or "Control Effectiveness %".
  • Mapping is not compliance: Linking an internal policy clause to a regulatory standard proves design intention; it does not demonstrate operational adherence.
  • Coverage is not effectiveness: Demonstrating that 100% of policy expectations exist in written text does not prove that systems enforce the policy or that personnel adhere to it.

Permissions Reference

ActionPermission KeyRequired Role(s)
View policy assessments, segments, and resultspolicy_assessment.readTenant Admin, Compliance Manager, Compliance Officer, Auditor
Create assessment, upload source, run extractionpolicy_assessment.runTenant Admin, Compliance Manager, Compliance Officer
Accept/reject mappings, assign results, complete assessmentpolicy_assessment.reviewTenant Admin, Compliance Manager

OrviQ Enterprise Governance, Risk & Compliance Platform