title: “Improvement Schema” description: “Machine-readable representation of AI governance improvement activities.”
Improvement Schema
The Improvement Schema provides the machine-readable structure for representing activities undertaken to improve AI governance capabilities, processes, controls, risk treatment, evidence practices, evaluation, and organizational outcomes. The schema supports traceability from an identified opportunity, deficiency, finding, or lesson through the resulting decision, action, implementation, verification, and evidence.Purpose
Continuous improvement is an integral part of the AIGO governance lifecycle. The Improvement Schema supports structured recording of: * improvement opportunities * identified deficiencies * findings * root causes * improvement decisions * corrective actions * preventive actions * responsible owners * implementation activities * target outcomes * verification * supporting evidence * closure The schema provides the structural representation of an improvement record. It does not prescribe the specific improvement methodology an organization must use.Improvement inputs
Improvement activities may originate from many governance sources. Potential inputs include: * risk assessments * control assessments * monitoring results * incidents * changes * assurance findings * management reviews * evidence deficiencies * evaluation findings * validation failures * maturity assessments * stakeholder feedback * recurring governance exceptions * lessons learned The originating governance record should remain traceable to the improvement activity where applicable.Improvement lifecycle
A practical improvement lifecycle is: observation → finding → decision → action → implementation → verification → closure The exact sequence may vary depending on the nature and significance of the improvement. Each significant improvement should maintain sufficient information to establish: * what was identified * why improvement was required * what decision was made * who is accountable * what action was required * what was implemented * how the result was verified * what evidence supports the conclusionCorrective and preventive improvement
Improvement activities may address an existing deficiency or seek to prevent a recurring or potential deficiency. Examples include:Corrective improvement
Addresses an identified problem or deficiency. Examples: * remediation of a control weakness * correction of incomplete evidence * resolution of an assurance finding * correction of an evaluation failure * response to an incident-related governance weaknessPreventive improvement
Addresses an emerging or potential weakness before it results in a significant deficiency. Examples: * strengthening monitoring * improving governance criteria * introducing additional controls * improving training * changing assessment methodology * improving evidence collection The applicable improvement classification should follow the organization’s governance requirements and the canonical schema definitions.Root cause and systemic improvement
Organizations should consider whether an identified issue is isolated or indicates a broader governance weakness. Repeated findings may indicate that remediation of individual instances is insufficient. For example, recurring: * control deficiencies * evidence gaps * approval exceptions * evaluation failures * monitoring issues * classification inconsistencies may indicate that the underlying governance process, control design, criteria, or implementation approach requires improvement. Improvement records can therefore support both instance-level remediation and systemic governance improvement.Improvement traceability
Improvement records should remain connected to the governance records that generated or are affected by the improvement. Relationships may include: * AI systems * risks * controls * assessments * approvals * monitoring * incidents * changes * assurance * evidence * management reviews * retirement activities A representative relationship is: governance observation → finding → improvement decision → action → evidence → verification This allows organizations to trace improvement activities back to their governance origin and forward to their resulting outcome.Improvement and evaluation
The AIGO machine-readable evaluation layer can provide additional inputs to improvement activities. Evaluation results may identify: * unmet governance conditions * missing information * control deficiencies * evidence gaps * approval conditions that require attention * recurring governance failures The architecture separates governance records from evaluation logic: governance records → rules → evaluation → result → improvement An evaluation result should therefore be treated as an input to governance decision-making rather than as an automatic determination of the required improvement response.Improvement and validation
Repository and machine-readable validation can also produce improvement inputs. Validation may identify: * schema inconsistencies * broken references * missing relationships * invalid rule definitions * invalid evaluation fixtures * orchestration configuration issues * document integrity problems These findings can be managed through appropriate improvement activities. A successful validation result indicates the state of the validated repository or artifact scope at the time of validation. It does not mean that the governance system is permanently complete or that an organization is legally compliant.Verification and closure
An improvement should not normally be considered complete solely because an action was marked as implemented. Where appropriate, the organization should verify whether the intended improvement outcome was achieved. Verification may include: * control reassessment * evidence review * monitoring results * testing * assurance activity * evaluation * management review Closure should preserve the evidence supporting the conclusion that the improvement was completed or otherwise resolved according to the applicable governance process.Machine-readable governance
The Improvement Schema forms part of the broader AIGO machine-readable governance layer. It can establish relationships with: * AI systems * risks * controls * assessments * approvals * monitoring * incidents * changes * assurance * evidence * management reviews * retirement This allows improvement to operate as a lifecycle-wide governance capability rather than as an isolated corrective-action record.Validation
Improvement records participate in the broader AIGO validation and traceability model. Repository validation may include checks relating to: * schema structure * references * traceability * evidence relationships * framework consistency * document integrity * repository health * machine-readable governance artifacts A validation result provides evidence about the integrity of the validated repository or artifact scope. It does not by itself establish legal compliance, certification, accreditation, or regulatory conformity.Canonical source
Machine-readable schema:schemas/12-improvement/12-AIGO-Improvement-Schema-v0.1.json
Schema documentation:
schemas/12-improvement/12-AIGO-Improvement-Schema-Documentation-v0.1.md