Impact Assessment
Purpose
Assess the positive and negative consequences of AI systems for affected people and organizations, and provide the findings to risk assessment and deployment decisions.
Scope
All AI systems Rygen builds or uses, including X1, Corsair and internal tools.
Assessment Triggers
Assess new systems and reassess after significant modifications, changes in deployment context or users, client onboarding with unique AI usage, and regulatory or jurisdictional changes. Confirm each assessment as current, or update it, at least annually during management review (AI-012).
Impact Categories
Identify who is affected and assess the consequences below. Use the standard template and state each consequence once, with cross-references where needed.
Individual Impacts
Identify direct users, people affected by outputs without using the system, and people whose data is processed. Consider efficiency, decision-making autonomy, skills and roles, work satisfaction and privacy.
Group Impacts
Identify effects on client organizations, Rygen teams and relevant professional and supply-chain roles. Consider operations, finances, resource burden, liability, reputation and competitive position. State the consequence for the group; likelihood, scoring and treatment decisions belong in AI-008.
Societal Impacts
Assess material second-order effects case by case: supply-chain efficiency, resource use and the environment, safety, industry practices and economic productivity. Explain limited or absent applicability rather than adding generic benefits. Safety assessment covers intended operation, foreseeable misuse, incorrect or unavailable output and the human safeguards before consequences take effect.
Assessment Process
Phase 1: System Context Analysis
Record purpose, AI capabilities, technical and societal context (business environment and user sophistication), intended uses and foreseeable misuse (intentional, unintentional and safety-critical failure). Identify jurisdictions where the system operates or processes data and applicable AI, privacy, industry and cross-border requirements. An unchanged inherited jurisdiction review may be cited by document and date instead of copied; state any additional processing location or difference. Keep dated findings and unresolved questions explicit.
Phase 2: Stakeholder Identification
Identify direct users, affected people and data subjects, client and internal teams, supply-chain roles and relevant societal groups. Use roles rather than individual names.
Phase 3: Consequence Assessment
For each stakeholder group, record positive and negative consequences. Describe what happens to whom, including harms from misuse or failure and any safety implications. Do not substitute generic categories or invented stakeholder views. Preserve open questions and their source; key findings select from the assessed consequences and add nothing new.
Phase 4: Documentation
Complete the standard template, review it with appropriate stakeholders, file it in the AIMS and supply it to AI-008. Record who receives the findings, what is communicated and when; explain any decision that no communication is needed.
Integration with Risk Assessment Process
Impact assessment is a mandatory input to AI-008. The impact assessment identifies consequences; the risk assessment records their evaluation and disposition.
Mandatory Evaluation Requirement
Every negative consequence must be evaluated under AI-008, using its likelihood and impact scales and treatment options. Record the risk or justified exclusion in the system risk assessment; do not drop a consequence because it appears minor.
Exclusion Justification Requirement
An exclusion requires a written reason in the risk assessment traceability matrix, reviewed and approved by the Principal AI Engineer and retained for audit. Permitted reasons are:
- Negligible risk: Very Low under AI-008 (Risk Levels & Required Action).
- Existing control coverage: identify the existing risk with adequate controls.
- Outside the AIMS scope: explain the scope boundary. This includes a general business risk that is not specific to AI, such as a delivery schedule or maintenance cost.
Information Flow to Risk Assessment
Negative consequences become candidate risks; consequence severity informs impact scoring; stakeholder context informs analysis. Positive consequences may inform opportunity assessment.
Traceability Requirements
A traceable disposition must be maintained for every negative impact.
Traceability Matrix
Keep the matrix only in the system risk assessment, under Impact Assessment Traceability. The impact assessment points to it; do not duplicate it.
Required Elements
For each negative consequence, record its reference (its impact ID, or the impact assessment section it comes from and the row where that section holds a table), description, risk-register ID if included, disposition (Included or Excluded), and justification for exclusion.
Revision History
| Version | Date | Author | Summary of Change |
|---|---|---|---|
| 1.0 | 2025-06-05 | Field Bradley | Initial draft. |
| 1.1 | 2025-09-02 | Field Bradley | Migrated to markdown and gitlab |
| 1.2 | 2025-09-10 | Field Bradley | Improved guidance for assessing individuals, groups, and society |
| 1.3 | 2026-01-13 | Field Bradley | Formalized impact-to-risk linkage with mandatory evaluation, exclusion justification requirements, and traceability matrix |
| 1.4 | 2026-01-13 | Field Bradley | Added explicit safety assessment as societal impact per ISO 42001 clauses 6.1.4 and A.5.5 |
| 1.5 | 2026-09-10 | Field Bradley | Impact disposition matrix relocated to the risk assessment (AI-008) so it is recorded once, not duplicated in both documents |
| 1.6 | 2026-11-02 | Field Bradley | Added a yearly confirmation of each assessment; exclusion justifications recorded in the risk assessment’s matrix, matching where the matrix is held; impacts may be referenced by section and row; management review verification assigned to top management, as the AI Governance Committee holds no regular meetings (AI-2173) |
| 1.7 | 2026-11-05 | Field Bradley | Matched the concise impact template, keeping assessment, safety, consultation and review requirements, and cites AI-008 for exclusions and treatment options (AI-2132, AI-2165, AI-2131). |
| 1.8 | 2026-11-10 | Field Bradley | Removed the management review verification of impact dispositions; each risk assessment records the disposition of every negative impact and is approved in its merge request (AI-1976). |