How it works · Prototype and research boundary
From source change to accountable decision
Each stage exposes its input, output, uncertainty, and human role. The current prototype covers analysis through impact review; action and closure remain research direction.
Inspectable claims. Declared scope. Human accountability.
The causal sequence
A change is not closed when analysis ends
Scroll normally through seven distinct stages. The emphasis follows the reading order; no content is pinned, replaced, or hidden.
Change
A controlled source changes
An approved version pair and declared portfolio establish what may be evaluated.
QMS-SOP-014 · v4.1 → v4.2Propagate
Meaning is separated from text
Exact changed evidence is normalized into a reviewable Semantic Delta.
OPTIONAL → MANDATORY · 30 → 15 daysInspect
Dependencies are evaluated
Candidate paths retain relationship, origin, method, evidence, version, and state.
Confirmed · inferred · blockedDecide
Uncertainty reaches a person
An accountable reviewer confirms, corrects, rejects, or leaves a finding unresolved.
TR-22 · human decision retained- Research direction
Act
Accepted findings become owned work
The intended workflow connects accepted findings to bounded actions and accountable owners.
M6 incomplete · not current capability - Research direction
Prove
Evidence is attached to exact actions
The intended record binds completion evidence, versions, decisions, and limitations.
M7 incomplete · not current capability - Research direction
Close
Closure follows declared gates
Only resolved decisions, completed or accepted actions, evidence, exceptions, and approvals can support closure.
M8 incomplete · synthetic target workflow
Synthetic case The sequence explains intended causality. It is not live analysis, and Act, Prove, and Close are not current M0–M4 product capabilities.
1. Fix the source pair and scope
Input: approved old and new source versions plus a declared portfolio. Output: immutable source snapshots and an explicit evaluation universe. Human role: resolve source identity or unsupported-format issues before analysis proceeds.
- Uncertainty stays visible: missing, stale, failed, or excluded sources do not enter unaffected scope.
2. Separate text change from meaning change
Input: bounded document regions. Output: typed semantic units and structural/semantic change records with exact old/new evidence. Human role: correct or leave ambiguous interpretations unresolved.
Text, meaning, and impact are different records
Move through three levels. Each level keeps its own evidence basis and does not silently become the next.
- Changed text exact evidence
- Semantic Delta reviewed interpretation
- Downstream impact dependency-based inference
Level 1 · exact evidence
Text diff
Version 4.1The Quality Investigator
mayinitiate the investigation within30 calendar daysof intake.
Version 4.2The Quality Investigator must initiate the investigation within 15 calendar days of intake.
This establishes: which exact source characters differ. It does not establish normalized meaning or downstream impact.
Level 2 · reviewed interpretation
Semantic Delta · SD-SYN-001
- Actor
- Quality Investigator
- Action
- initiate investigation
- Modality
OPTIONAL→ MANDATORY- Timing
WITHIN 30 CALENDAR DAYS→ WITHIN 15 CALENDAR DAYS
This establishes: the normalized meaning accepted for this synthetic narrative. It does not establish which downstream objects depend on that meaning.
Level 3 · dependency-based inference
Downstream impact
| Object | Relationship state | Bounded outcome |
|---|---|---|
| WI-08 | confirmed | potentially affected |
| CL-04 | confirmed | potentially affected |
| TR-22 | inferred — review required | review required |
| FRM-11 | evaluated | no relevant impact found within evaluated scope |
| SOP-21 | blocked | not evaluated |
This establishes: outcomes supported by eligible paths inside the declared evaluation. Text comparison alone did not produce these findings.
Declared boundary: A result applies only to this declared synthetic portfolio, frozen graph, applicability policy, and evaluation budget.
3. Establish possible dependency paths
Input: confirmed changes and a versioned candidate graph. Output: candidates that retain relationship type, origin, method, evidence, state, version, and coverage. Human role: review inferred or policy-sensitive relationships.
Inspect the path, not just the outcome
Select a target to inspect its node, relationship basis, evidence, lifecycle state, version, and bounded impact reason.
Choose a target
Target node · WI-08
Investigation Work Instruction
- Source evidenceQMS-SOP-014 v4.2
- Semantic DeltaSD-SYN-001
- Relationshipreferences and restates
- TargetWI-08
- Object
- WI-08 · Investigation Work Instruction
- Type
- Controlled work instruction
- State
- confirmed
- Reason
- The work instruction repeats both changed fields.
- Source
- QMS-SOP-014, Section 6.2, paragraph 2
- Evidence
- Section 2 cites QMS-SOP-014; section 4.1 repeats the 30-day initiation window.
- Relationship
- references and restates
- Origin
- exact reference plus structural match
- Method
- deterministic reference rule v1
- Lifecycle state
- confirmed
- Graph version
- GRAPH-SYN-2026-09-15-01
- Comparison
- CMP-SYN-2026-09-15-01
Target node · CL-04
Investigation Intake Checklist
- Source evidenceQMS-SOP-014 v4.2
- Semantic DeltaSD-SYN-001
- Relationshipimplements
- TargetCL-04
- Object
- CL-04 · Investigation Intake Checklist
- Type
- Controlled checklist
- State
- confirmed
- Reason
- The acceptance criterion contains the previous deadline.
- Source
- QMS-SOP-014, Section 6.2, paragraph 2
- Evidence
- Checklist item 7 asks whether initiation occurred within 30 calendar days.
- Relationship
- implements
- Origin
- field-to-claim match
- Method
- policy rule v3
- Lifecycle state
- confirmed
- Graph version
- GRAPH-SYN-2026-09-15-01
- Comparison
- CMP-SYN-2026-09-15-01
Target node · TR-22
Investigator Refresher Training
- Source evidenceQMS-SOP-014 v4.2
- Semantic DeltaSD-SYN-001
- Relationshiptrains
- TargetTR-22
- Object
- TR-22 · Investigator Refresher Training
- Type
- Training module
- State
- inferred — review required
- Reason
- The content is semantically related, but the relationship is not yet confirmed.
- Source
- QMS-SOP-014, Section 6.2, paragraph 2
- Evidence
- Module 3 teaches investigation initiation timing but does not cite the SOP identifier.
- Relationship
- trains
- Origin
- semantic candidate
- Method
- bounded candidate classifier v2
- Lifecycle state
- inferred — review required
- Graph version
- GRAPH-SYN-2026-09-15-01
- Comparison
- CMP-SYN-2026-09-15-01
Declared boundary: A result applies only to this declared synthetic portfolio, frozen graph, applicability policy, and evaluation budget.
4. Produce bounded impact findings
Input: a confirmed change, compatible frozen graph, applicability rules, and traversal budgets. Output: affected, potentially affected, review-required, bounded-unaffected, blocked, and not-evaluated states. Human role: decide materiality and context where policy requires it.
Inspect the path, not just the outcome
Select a target to inspect its node, relationship basis, evidence, lifecycle state, version, and bounded impact reason.
Choose a target
Target node · WI-08
Investigation Work Instruction
- Source evidenceQMS-SOP-014 v4.2
- Semantic DeltaSD-SYN-001
- Relationshipreferences and restates
- TargetWI-08
- Object
- WI-08 · Investigation Work Instruction
- Type
- Controlled work instruction
- State
- confirmed
- Reason
- The work instruction repeats both changed fields.
- Source
- QMS-SOP-014, Section 6.2, paragraph 2
- Evidence
- Section 2 cites QMS-SOP-014; section 4.1 repeats the 30-day initiation window.
- Relationship
- references and restates
- Origin
- exact reference plus structural match
- Method
- deterministic reference rule v1
- Lifecycle state
- confirmed
- Graph version
- GRAPH-SYN-2026-09-15-01
- Comparison
- CMP-SYN-2026-09-15-01
Target node · CL-04
Investigation Intake Checklist
- Source evidenceQMS-SOP-014 v4.2
- Semantic DeltaSD-SYN-001
- Relationshipimplements
- TargetCL-04
- Object
- CL-04 · Investigation Intake Checklist
- Type
- Controlled checklist
- State
- confirmed
- Reason
- The acceptance criterion contains the previous deadline.
- Source
- QMS-SOP-014, Section 6.2, paragraph 2
- Evidence
- Checklist item 7 asks whether initiation occurred within 30 calendar days.
- Relationship
- implements
- Origin
- field-to-claim match
- Method
- policy rule v3
- Lifecycle state
- confirmed
- Graph version
- GRAPH-SYN-2026-09-15-01
- Comparison
- CMP-SYN-2026-09-15-01
Target node · TR-22
Investigator Refresher Training
- Source evidenceQMS-SOP-014 v4.2
- Semantic DeltaSD-SYN-001
- Relationshiptrains
- TargetTR-22
- Object
- TR-22 · Investigator Refresher Training
- Type
- Training module
- State
- inferred — review required
- Reason
- The content is semantically related, but the relationship is not yet confirmed.
- Source
- QMS-SOP-014, Section 6.2, paragraph 2
- Evidence
- Module 3 teaches investigation initiation timing but does not cite the SOP identifier.
- Relationship
- trains
- Origin
- semantic candidate
- Method
- bounded candidate classifier v2
- Lifecycle state
- inferred — review required
- Graph version
- GRAPH-SYN-2026-09-15-01
- Comparison
- CMP-SYN-2026-09-15-01
Declared boundary: A result applies only to this declared synthetic portfolio, frozen graph, applicability policy, and evaluation budget.
5. Research direction: act, prove, and close
Intended input: accepted impact decisions. Intended output: owned actions, implementation evidence, approvals, limitations, and a frozen Change Evidence Package. Human role: perform work, accept evidence, approve exceptions, and authorize closure. These M6–M8 capabilities are not current product availability.
Analysis is not closure
This synthetic target-workflow sequence shows why an impact result remains open until decisions, actions, evidence, scope exceptions, approvals, and the release basis satisfy explicit gates.
Negative example
Analysis complete; change still open
2 complete · 3 blocked · 2 waiting
- CompleteAnalysis complete
Bounded findings and exclusions recorded.
- CompleteBlocking decisions
2 / 2 decisions resolved.
- BlockedRequired actions
2 / 3 actions dispositioned; TR-22 remains open.
- BlockedImplementation evidence
Training completion evidence is missing.
- BlockedCoverage exception
SOP-21 is stale and neither evaluated nor accepted as an exception.
- WaitingRequired approvals
Approval cannot precede resolved evidence and scope gates.
- WaitingRelease basis
The comparison and evidence package are not frozen for closure.
Blocking reason: training evidence is missing and SOP-21 remains outside the evaluated basis.
Illustrative final state
Every declared gate is satisfied
7 / 7 gates complete
- CompleteAnalysis complete
Bounded findings and exclusions recorded.
- CompleteBlocking decisions
2 / 2 decisions resolved.
- CompleteRequired actions
3 / 3 actions completed, waived, or accepted as exceptions.
- CompleteImplementation evidence
Evidence is attached to each exact action and version.
- CompleteCoverage exception
SOP-21 is evaluated or explicitly accepted as a scoped exception.
- CompleteRequired approvals
Accountable approval records are present.
- CompleteRelease basis
Comparison, graph, decisions, actions, evidence, and limitations are frozen.
Boundary: Closed describes this synthetic target-workflow narrative only, not current product capability or a customer result.
Next step