5.6: SUP.10 Change Request Management

Process Definition

Purpose

SUP.10 Purpose: To ensure that change requests are managed, tracked, and controlled.

Change Request Management is a cornerstone of configuration-controlled development in safety-critical embedded systems. In ASPICE 4.0, SUP.10 governs how proposed modifications to baselined work products -- requirements, architecture, code, test cases, documentation -- are formally recorded, analyzed for impact, approved by competent authority, implemented under controlled conditions, and verified before closure. Without disciplined change management, traceability erodes, baselines become unreliable, and compliance evidence fragments.

AI integration elevates SUP.10 from a largely administrative burden into a proactive, intelligence-driven process. Rather than relying solely on human reviewers to mentally trace every dependency when a change request arrives, AI can instantly analyze the full traceability web, flag affected work products, estimate effort with calibrated models, and present the Change Control Board (CCB) with structured decision packages. The human retains full decision authority; the AI accelerates and deepens the analysis that supports that decision.

Outcomes

Base Practices with AI Integration

CCB Decision Authority: AI provides impact analysis and recommendations; the Change Control Board (CCB) retains final approval authority and may override AI recommendations based on business factors, resource constraints, or strategic considerations not captured in automated impact analysis. Human judgment is essential for balancing technical feasibility against project priorities.

Change Request Lifecycle

A change request progresses through a defined set of states. Each transition has explicit entry criteria, and AI assists at every stage by validating completeness, flagging anomalies, and accelerating analysis.

States and Transitions

Transition Rules

Note: Backward transitions (e.g., from Approved back to Under Analysis) are permitted when new information surfaces, but must be documented with justification. AI flags any backward transition for mandatory human review.

AI-Powered Impact Analysis

Impact analysis is the highest-value AI application within SUP.10. When a change request arrives, the AI traverses the project's traceability graph to identify every work product that may be affected, estimates the ripple effects, and quantifies the effort and risk.

Analysis Dimensions

Impact Analysis Report Structure

AI Change Impact Analysis
Change Request: CR-2025-015
Title: Add temperature compensation for GPIO timing

The diagram below visualizes the AI-generated impact analysis for this change request, showing affected components, traceability links, and estimated effort across hardware and software domains.

The AI generates a structured impact report that feeds directly into the CCB decision package:

# AI-Generated Impact Analysis Report (illustrative)
impact_analysis:
  change_request: CR-2025-015
  generated_by: AI Impact Analyzer v2.1
  generation_date: 2025-01-15
  confidence_score: 0.87

  requirements_impact:
    direct:
      - id: SWE-BCM-103
        description: "GPIO drive strength configuration"
        change_needed: "Add temperature parameter to drive strength lookup"
      - id: SWE-BCM-120
        description: "Timing compliance at temperature extremes"
        change_needed: "Update acceptance criteria for -40C operation"
    transitive:
      - id: SYS-BCM-045
        description: "System timing budget"
        change_needed: "Review system-level timing allocation"
    total_requirements_affected: 3

  architecture_impact:
    components_affected:
      - name: "GPIO Driver"
        coupling: high
        change_type: "Interface modification"
      - name: "Temperature Monitor Service"
        coupling: medium
        change_type: "New dependency"
    interface_changes: 1

  code_impact:
    files_modified:
      - path: "src/driver/gpio_driver.c"
        functions_affected: ["GPIO_SetDriveStrength", "GPIO_Init"]
        lines_estimated: 45
      - path: "src/driver/gpio_driver.h"
        functions_affected: ["GPIO_SetDriveStrength signature"]
        lines_estimated: 5
      - path: "src/service/temp_monitor.c"
        functions_affected: ["TempMonitor_GetCompensation"]
        lines_estimated: 30
    total_lines_estimated: 80

  test_impact:
    unit_tests:
      - test_gpio_drive_strength.c (modify)
      - test_temp_compensation.c (new)
    integration_tests:
      - SWE-IT-GPIO-001 (re-execute)
      - SWE-IT-BCM-003 (re-execute)
    system_tests:
      - SYS-ST-TIMING-001 (re-execute at temperature extremes)
    total_tests_affected: 5

  effort_estimate:
    implementation: 16 hours
    unit_testing: 8 hours
    integration_testing: 4 hours
    documentation: 4 hours
    total: 32 hours
    confidence: medium

  risk_assessment:
    overall_risk: low
    factors:
      - "GPIO interface change affects one downstream consumer"
      - "Temperature compensation is additive, not modifying existing logic"
      - "Existing test infrastructure covers temperature sweep"
    mitigations:
      - "Boundary value testing at -40C, +25C, +85C"
      - "Regression test full GPIO suite after change"

Confidence Scoring

AI impact analysis includes a confidence score (0.0 to 1.0) that reflects the completeness and quality of the traceability data available. Low confidence scores signal to the CCB that manual analysis should supplement the AI output.

AI-Assisted Change Management

The following diagram traces the complete change request lifecycle, from submission and AI-assisted impact analysis through CCB review, implementation, and verification.

Change Classification

Consistent classification of change requests is essential for prioritization, routing, and metrics. AI assists by analyzing the change description, affected components, and historical patterns to suggest both type and severity classifications. Human reviewers confirm or override the AI suggestion.

Type Classification

Severity Classification

AI Classification Workflow

1. Text Analysis: AI parses the CR title, description, and justification fields using NLP to extract intent keywords and affected domain terminology.
2. Artifact Linkage: AI examines linked problem reports, requirements, and risk register entries to infer type (e.g., a CR linked to a PR is likely Corrective).
3. Component Analysis: AI maps the affected files or requirements to safety classification (ASIL level) and architectural tier (driver, service, application) to inform severity.
4. Historical Comparison: AI compares against previously classified CRs with similar characteristics to calibrate the suggestion.
5. Human Confirmation: The suggested classification is presented to the triage owner, who confirms or overrides with documented rationale.

Traceability Through Changes

Bidirectional traceability is an ASPICE core requirement (O4) and becomes especially challenging when changes ripple across multiple work product levels. AI maintains traceability integrity throughout the change lifecycle.

Traceability Dimensions

Traceability Integrity Checks

AI performs continuous traceability integrity checks and reports anomalies:

Suspect Link Management

When a change request modifies a work product, all traceability links originating from or terminating at that work product become "suspect" until reviewed. AI manages this process:

1. Automatic Flagging: When a CR moves to "In Implementation," AI marks all downstream links as suspect.
2. Review Assignment: AI generates a review checklist for each suspect link, assigned to the appropriate work product owner.
3. Resolution Tracking: As owners confirm or update links, AI clears suspect flags and records the review evidence.
4. Completeness Gate: Before a CR can move to "Closed," AI verifies that zero suspect links remain.

Change Request Template

# Change Request (illustrative example)
change_request:
  id: CR-(year)-(number)
  title: "Add temperature compensation for GPIO timing"
  status: approved
  created: (creation date)
  requester: SW Lead

  description: |
    Implement temperature-dependent GPIO drive strength configuration
    to maintain timing compliance at cold temperatures (-40°C).

  justification: |
    Problem PR-2025-042 identified timing violation at cold temperature.
    This change addresses the root cause by compensating for temperature-
    dependent transistor switching characteristics.

  classification:
    type: corrective
    severity: medium
    ai_suggested_type: corrective
    ai_suggested_severity: medium
    ai_confidence: 0.91
    human_override: false

  related_items:
    problem: PR-2025-042
    requirements:
      - SWE-BCM-103
      - SWE-BCM-120
    risks:
      - RSK-003

  impact_analysis:
    ai_generated: true
    requirements_impact: medium
    code_impact: low
    test_impact: medium
    effort_estimate: 24 hours
    risk: low

  ccb_review:
    date: 2025-01-17
    attendees:
      - Project Manager
      - SW Lead
      - QA Lead
      - System Architect
    decision: approved
    conditions:
      - "Add tests for boundary temperatures"
      - "Update design document"

  implementation:
    assigned_to: SW Developer
    target_date: 2025-01-22
    actual_date: 2025-01-21
    work_items:
      - WI-2025-089 (Implementation)
      - WI-2025-090 (Unit tests)
      - WI-2025-091 (Documentation)

  verification:
    status: complete
    tests_passed: true
    regression_passed: true

  closure:
    date: 2025-01-22
    closed_by: SW Lead

Tool Integration

SUP.10 effectiveness depends on tight integration between the change management tool, version control, requirements management, and CI/CD systems. The following table maps common ALM tools to SUP.10 capabilities and describes how AI enhances each integration point.

Tool-to-Process Mapping

Integration Architecture

A typical AI-enhanced change management integration connects the following systems:

Configuration Example: Jira + GitLab + AI

# Integration configuration (illustrative)
change_management_integration:
  cr_tracker:
    tool: jira
    project_key: BCM
    issue_type: "Change Request"
    custom_fields:
      ai_impact_score: customfield_10100
      ai_classification: customfield_10101
      ai_confidence: customfield_10102
    workflow:
      states: [Draft, Submitted, Under Analysis, Analyzed, Approved,
               In Implementation, Verification, Closed, Rejected, Deferred]
      transitions:
        - from: Submitted
          to: Under Analysis
          trigger: auto  # AI triggers analysis on submission
        - from: Under Analysis
          to: Analyzed
          trigger: ai_analysis_complete
        - from: Approved
          to: In Implementation
          trigger: manual  # Human creates work items

  version_control:
    tool: gitlab
    branch_naming: "cr/{cr_id}/{short_description}"
    commit_convention: "CR-{id}: {description}"
    merge_request_template: |
      ## Change Request
      CR: {cr_id}
      Impact Level: {ai_impact_level}
      Affected Requirements: {requirement_list}

      ## Verification Checklist
      - [ ] Unit tests updated
      - [ ] Integration tests re-executed
      - [ ] Documentation updated
      - [ ] Traceability matrix verified

  ai_service:
    endpoint: "https://ai-analysis.internal/api/v2"
    triggers:
      - event: cr_submitted
        action: run_impact_analysis
      - event: cr_work_items_complete
        action: verify_traceability_completeness
      - event: baseline_requested
        action: check_open_crs_against_baseline

  notifications:
    channels:
      - type: email
        recipients: ccb_members
        events: [cr_analyzed, cr_approved, cr_rejected]
      - type: teams_channel
        channel: "#bcm-changes"
        events: [cr_submitted, cr_closed, cr_emergency]

HITL Protocol for Changes

Human-in-the-Loop (HITL) controls are non-negotiable for SUP.10 compliance. AI accelerates analysis and automates routine checks, but humans own every decision that affects baselined work products.

HITL Gate Definitions

Escalation Rules

Override Documentation

When a human overrides an AI recommendation (e.g., changes the suggested classification or rejects the impact estimate), the override must be documented:

# Override record (illustrative)
override_record:
  cr_id: CR-2025-022
  field_overridden: severity
  ai_recommendation: medium
  human_decision: high
  rationale: |
    AI did not account for upcoming regulatory audit in Q2.
    Unresolved timing issue during audit would constitute
    a non-conformance finding. Elevating severity to ensure
    resolution before audit window.
  overridden_by: QA Lead
  date: 2025-02-10

Metrics and KPIs

Effective change management requires quantitative monitoring. The following metrics track both process efficiency and AI contribution quality.

Process Efficiency Metrics

AI Effectiveness Metrics

Reporting Cadence

Work Products

Implementation Checklist

Use this checklist when establishing or improving SUP.10 Change Request Management with AI integration.

Process Setup

• Define CR workflow states and transitions in ALM tool matching the lifecycle defined in this chapter
• Configure mandatory fields for CR creation: title, description, justification, requester, affected components
• Establish CR numbering convention (e.g., CR-YYYY-NNN) with automatic assignment
• Define type and severity classification criteria and configure in ALM tool
• Set SLA targets for each severity level and configure aging alerts
• Charter the Change Control Board: membership, quorum rules, meeting cadence, decision recording procedure

AI Integration

• Deploy AI impact analysis service connected to requirements management tool and version control
• Configure traceability data feeds so AI can traverse requirement-to-code-to-test links
• Set up duplicate detection using historical CR corpus and NLP similarity matching
• Calibrate effort estimation model using at least 20 historical CRs with actual effort data
• Define confidence score thresholds and corresponding human review requirements
• Implement automated classification suggestion with human confirmation workflow

Tool Integration

• Connect ALM tool to version control via API (branch naming, commit linking, merge request correlation)
• Configure CI/CD pipeline to trigger impact analysis on merge request creation referencing a CR
• Set up webhook-based notifications for CR state transitions to relevant stakeholders
• Create dashboard views: CR pipeline, aging report, SLA compliance, AI effectiveness metrics
• Configure baseline impact checks that prevent baseline creation with unresolved CRs

HITL and Governance

• Document HITL gates (G1-G6) in the project's change management plan
• Define escalation rules for SLA breaches, low-confidence AI outputs, and emergency CRs
• Create override documentation template and train team on when and how to override AI suggestions
• Establish CCB decision recording procedure (attendees, decision, conditions, rationale)
• Define backward transition rules and mandatory justification requirements

Metrics and Continuous Improvement

• Instrument ALM tool to capture all metrics defined in the Metrics and KPIs section
• Schedule monthly metrics review with process owner
• Establish AI effectiveness baseline during first quarter of operation
• Define improvement targets for AI accuracy and process efficiency after baseline period
• Plan quarterly retrospective on change management process effectiveness

Summary

SUP.10 Change Request Management:

• AI Level: L2 (AI analysis, human approval)
• Primary AI Value: Impact analysis, effort estimation, classification, traceability maintenance, duplicate detection
• Human Essential: CCB decision authority, implementation sign-off, verification confirmation, override decisions
• Key Outputs: CR records, impact analysis reports, traceability updates, CCB decision packages, change metrics
• Integration: Connects ALM, requirements management, version control, CI/CD, and notification systems
• HITL Protocol: Six defined gates (G1-G6) ensuring human authority at every critical decision point