AI-BOK Reference Architecture AI-BOK Reference Architecture (v1.2, June 2026) ================================================ This ArchiMate model is the architecture translation of the AI Body of Knowledge (AI-BOK) v1.2, written by Jan Willem van Veen (ArchiXL). The model provides a complete reference framework for AI governance and architecture, structured according to the ArchiMate 3.2 metamodel across all layers, including the cognition plane alongside the data, control and management planes. Elements: 712 Relations: 1113 Views: 52 (folders 01-09: overview, business & organisation, application & technology, motivation & principles, maturity & quality, 13 knowledge areas, failure modes, regulatory landscape, GEMMA pattern) Source references per element are kept in dct:source properties (AI-BOK v1.2, module + section). Full edition, templates and toolkit: https://www.ai-bok.nl (models: /architecture). The accompanying paper was peer-reviewed at EA4AI @ CBI-EDOC 2026 (Springer LNBIP). KA1: AI Governance Create conditions under which AI systems deliver maximum value within acceptable risk boundaries. High 3-Defined Very High High AI-BOK v1.2, Module 1, section 1.1 KA2: AI Strategy & Portfolio Management Maximum organisational value from AI investments through systematic selection and prioritisation. High 3-Defined Medium Limited AI-BOK v1.2, Module 1, section 2.1 KA3: AI Architecture A coherent, repeatable and evolvable blueprint for AI capabilities. Very High 3-Defined High Limited AI-BOK v1.2, Module 1, section 3.1 KA4: AI Lifecycle Management Predictable, repeatable and governable progress of AI initiatives. High 3-Defined Medium Limited AI-BOK v1.2, Module 1, section 4.1 KA5: Data & Semantics for AI Reliable, traceable and semantically anchored data foundation for AI. Very High 3-Defined High High AI-BOK v1.2, Module 1, section 5.1 KA6: Model Management Reliable, reproducible and responsible model portfolio. Very High 3-Defined High High AI-BOK v1.2, Module 1, section 6.1 KA7: AI Interaction & User Experience AI interactions that are user-centred, reliable and inclusive. High 2-Repeatable Medium Limited AI-BOK v1.2, Module 1, section 7.1 KA8: AI Operations Reliable, observable and cost-efficient AI operations in production. High 3-Defined High Limited AI-BOK v1.2, Module 1, section 8.1 KA9: AI Risk Management & Safety Proactively identify, assess and mitigate AI-related risks. Very High 3-Defined Very High High AI-BOK v1.2, Module 1, section 9.1 KA10: AI Compliance & Audit Verifiable, risk-based compliance of AI systems. High 3-Defined Very High High AI-BOK v1.2, Module 1, section 10.1 KA11: AI Ethics & Responsible AI Ethical considerations structurally embedded in the design and deployment of AI. High 2-Repeatable High High AI-BOK v1.2, Module 1, section 11.1 KA12: AI Knowledge & Context Management AI systems produce grounded, traceable and semantically vamemberated output. Very High 3-Defined High High AI-BOK v1.2, Module 1, section 12.1 KA13: AI Literacy & Capability Development Build and sustain the human knowledge, skills and judgement required to develop, deploy, supervise, use and decommission AI systems responsibly. Treats AI literacy as a structural capability tied to systems in production, with role-specific competencies, refresh cycles, evidence and measurement. Anchors EU AI Act Article 4 compliance. High AI-BOK v1.1 Assurance & Accountability AI-BOK v1.1, Module 1, KA13 Idee & Kans Identify AI opportunities and assess business value. Design Architecture, data preparation and model selection. Development Training, fine-tuning, prompt engineering and evaluation. Production Deployment, integration, monitoring and operations. Value Measurable business value and continuous improvement. AI Governance & Steering Establishing and enforcing AI policy, principles and mandates. High AI-BOK v1.2, Module 1, section 1.3 AI Strategy & Portfolio Management Selection, prioritisation and monitoring of AI initiatives. High AI-BOK v1.2, Module 1, section 2.3 AI Architecture Management Designing and maintaining the AI reference architecture. Very High AI-BOK v1.2, Module 1, section 3.3 AI Lifecycle Management Guiding AI initiatives through all phases of the lifecycle. High AI-BOK v1.2, Module 1, section 4.3 AI Data Management & Semantics Ensuring data quality, semantic precision and knowledge sources. Very High AI-BOK v1.2, Module 1, section 5.3 AI Model Management Registration, validation and version management of AI models. Very High AI-BOK v1.2, Module 1, section 6.3 AI Interaction Management Design and optimisation of AI user interactions. High AI-BOK v1.2, Module 1, section 7.3 AI Operations Management Deployment, monitoring and operational reliability. High AI-BOK v1.2, Module 1, section 8.3 AI Risk Management Identification, assessment and mitigation of AI risks. Very High AI-BOK v1.2, Module 1, section 9.3 AI Compliance & Audit Ensuring compliance with laws and regulations. High AI-BOK v1.2, Module 1, section 10.3 AI Ethics & Accountability Ensuring ethical principles in AI systems. High AI-BOK v1.2, Module 1, section 11.3 AI Knowledge Management Curation and governance of knowledge sources for AI grounding. Very High AI-BOK v1.2, Module 1, section 12.3 AI Literacy Management AI Portfolio Overview of all AI initiatives with status and priority. No personal data Confidential 5 years na afloop AI-BOK v1.2, Module 1, section 2.4 AI Business Case Justification of AI initiative with expected value and risks. No personal data Confidential 5 years na afloop AI-BOK v1.2, Module 3, section 3.1.1 Prompt Library Curated collection of prompt templates and strategies. Possibly personal data Internal AI-BOK v1.2, Module 1, section 7.4 AI Risk Register Register of identified AI risks with mitigation. Possibly personal data Confidential Permanent AI-BOK v1.2, Module 1, section 9.4 Algorithm Register Public register of deployed AI systems (transparency). No personal data Public Permanent (legal obligation) Mandatory (Art. 71) AI-BOK v1.2, Module 1, section 10.4 Knowledge Source Register Register of authoritative knowledge sources with classification. No personal data Internal AI-BOK v1.2, Module 1, section 12.4 Shadow AI Inventory Register of unapproved AI applications in the organisation. Possibly personal data Confidential AI-BOK v1.2, Module 3, section 3.1 AI Strategy Document Multi-year AI strategy document, reviewed annually. No personal data Internal AI-BOK v1.2, Module 1, section 2.4 AI Roadmap Rolling 12-18 month roadmap for AI initiatives. No personal data Internal AI-BOK v1.2, Module 1, section 2.4 AI Capacity Plan Plan for talent, technology, data and budget. No personal data Confidential AI-BOK v1.2, Module 1, section 2.6 AI Technology Radar Semi-annual overview: adopt, trial, assess, hold. No personal data Internal AI-BOK v1.2, Module 1, section 3.6 Integration Patterns Catalogue Catalogue of integration patterns (MCP, A2A, governance hooks). No personal data Internal AI-BOK v1.2, Module 1, section 3.13 Interaction Patterns Catalogue Catalogue of AI interaction patterns and guidelines. No personal data Internal AI-BOK v1.2, Module 1, section 7.4 Trusted Knowledge Sources Catalogue Register with authority levels and metadata per knowledge source. No personal data Internal AI-BOK v1.2, Module 1, section 12.4 Interaction Een uitwisseling tussen AI-systeem en gebruiker of ander systeem. AI-BOK v1.2, Module 3, section 3.2 Decision Uitkomst van een AI-ondersteund of -uitgevoerd besluitvormingsproces. AI-BOK v1.2, Module 3, section 3.2 AI Policy Framework Organisation-wide AI policy with principles, roles and mandates. No personal data Internal Permanent AI-BOK v1.2, Module 1, section 1.2 Agent Mandate Formal specification of autonomy boundaries and escalation rules per agent. No personal data Internal Agent lifetime AI-BOK v1.2, Module 1, section 1.5 AI Governance Charter Formal document with AI policy, principles, roles and mandates. No personal data Internal AI-BOK v1.2, Module 3, section 3.1 Data Contract Agreement on data quality and delivery between systems. Possibly personal data Confidential AI-BOK v1.2, Module 3, section 3.1 AI Reference Architecture Blueprint for AI capabilities including Cognition Plane. No personal data Internal Permanent AI-BOK v1.2, Module 1, section 3.4 Terminology Framework Authoritative concept framework (SKOS/NL-SBB) for AI systems. No personal data Public AI-BOK v1.2, Module 1, section 5.4 Model Card Standardised documentation of an AI model. Possibly personal data (training data description) Internal Model lifetime + 5 years Mandatory for high-risk AI AI-BOK v1.2, Module 1, section 6.4 Dataset Specification Documentation of dataset per Datasheets for Datasets. Possibly personal data (dataset may contain PII) Internal Dataset lifetime + 5 years AI-BOK v1.2, Module 1, section 5.13 EU AI Act Compliance Checklist Checklist for EU AI Act compliance per risk class. No personal data Internal Directly related AI-BOK v1.2, Module 3, section 3.1 Red Team Report Results of adversarial testing of AI systems. Possibly personal data Secret 5 years Recommended AI-BOK v1.2, Module 3, section 3.1 Bias Audit Report Results of bias analysis and fairness assessment. Possibly special category personal data (bias on protected groups) Confidential 10 years Mandatory for high-risk AI (Art. 10) AI-BOK v1.2, Module 3, section 3.1 AI Bill of Materials Complete inventory of components, models, data and dependencies. No personal data Internal Recommended AI-BOK v1.2, Module 3, section 3.1 Architecture Decision Record Documentation of architecture decisions with rationale. No personal data Internal AI-BOK v1.2, Module 3, section 3.1 Experiment Log Documentation of ML experiments with parameters and results. Possibly personal data Internal 5 years AI-BOK v1.2, Module 3, section 3.1 Decommissioning Report Documentation of responsible decommissioning of AI systems. No personal data Internal 5 years AI-BOK v1.2, Module 3, section 3.1 RACI Matrix for AI Roles Matrix documenting responsibilities per AI role and KA. No personal data Internal AI-BOK v1.2, Module 2, section 2.3 Architecture Conformity Report Report on conformity to the AI reference architecture. No personal data Internal AI-BOK v1.2, Module 1, section 3.8 Retraining Plan Per model: frequency, validation criteria, approval process. No personal data Internal AI-BOK v1.2, Module 1, section 4.6 Quality Gate Checklist Checklist per lifecycle phase with criteria and deliverables. No personal data Internal AI-BOK v1.2, Module 3, section 3.1 Decommissioning Protocol Protocol for responsible decommissioning of AI systems. No personal data Internal AI-BOK v1.2, Module 3, section 3.1.8 Data Lineage Documentation Documentation of data provenance and transformations. Possibly personal data Confidential AI-BOK v1.2, Module 1, section 5.6 Data Quality Report Report on data quality per dataset. Possibly personal data Internal AI-BOK v1.2, Module 1, section 5.4 Human-in-the-Loop Protocol Protocols and escalation specifications per risk level. No personal data Internal AI-BOK v1.2, Module 1, section 7.6 Drift Detection Report Rapport bij gedetecteerde data- of conceptdrift. Possibly personal data Internal AI-BOK v1.2, Module 1, section 8.6 AI Incident Report Incidentrapport met root-cause-analyse. Possibly personal data Confidential AI-BOK v1.2, Module 1, section 8.8 AI Incident Response Plan Plan voor AI-specifieke incidentrespons. No personal data Confidential AI-BOK v1.2, Module 1, section 9.6 AI Ethics Framework Organisation-specific ethical framework and code of conduct for AI. No personal data Internal AI-BOK v1.2, Module 1, section 11.4 Explainability Documentation Per AI system and decision type. No personal data Internal AI-BOK v1.2, Module 1, section 11.6 Knowledge Currency Report Report on currency and refresh schedule of knowledge sources. No personal data Internal AI-BOK v1.2, Module 1, section 12.8 Lessons Learned Document Documentation of lessons learned during evolution or decommissioning. No personal data Internal AI-BOK v1.2, Module 3, section 3.1.8 AI Board Strategic decision-making. AI-BOK v1.2, Module 2, section 2.5 AI Review Committee Tactical quality gate reviews. AI-BOK v1.2, Module 2, section 2.5 AI Ethics Committee Advisory on ethical dilemmas. AI-BOK v1.2, Module 2, section 2.5 AI Risk Committee Risk overview and acceptance. AI-BOK v1.2, Module 2, section 2.5 AI Centre of Excellence Central knowledge centre for AI expertise, best practices and reusable components. AI-BOK v1.2, Module 2, section 2.5 OpenAI American AI company, creator of GPT-4 and ChatGPT. https://www.openai.com LLM Provider Commercial SaaS US High Yes Anthropic AI-veiligheidsbedrijf, maker van Claude. https://www.anthropic.com LLM Provider Commercial SaaS US High Yes Meta Technologiebedrijf, maker van open-source LLaMA-modellen. https://ai.meta.com LLM Provider Open Source US High Yes Google Technology company, creator of Gemini and TensorFlow. https://ai.google LLM Provider Commercial SaaS US High Yes Google Cloud Cloud AI-platform met Vertex AI. https://cloud.google.com/vertex-ai Cloud Platform Commercial SaaS US High Yes Google PAIR People + AI Research initiative. https://pair.withgoogle.com Research Open Source US Medium Nee Cohere Enterprise AI platform for NLP and embeddings. https://www.cohere.com LLM Provider Commercial SaaS Canada High Yes Microsoft Technologiebedrijf, Azure AI en enterprise-integratie. https://azure.microsoft.com/ai Cloud Platform Commercial SaaS US High Yes Amazon (AWS) Cloud provider with Bedrock, SageMaker and AI services. https://aws.amazon.com/ai Cloud Platform Commercial SaaS US High Yes NVIDIA GPU- en AI-infrastructuurleverancier. https://www.nvidia.com/ai Hardware & AI Infra Commercial US High Yes LangChain Inc. Framework for LLM applications, RAG and agents. https://www.langchain.com AI Framework Open Source US Medium Nee CrewAI Multi-agent orkestratie framework. https://www.crewai.com AI Framework Open Source US Low Nee LlamaIndex Data framework for LLM applications and RAG. https://www.llamaindex.ai AI Framework Open Source US Medium Nee Pinecone Managed vector database voor AI-applicaties. https://www.pinecone.io Vector Database Commercial SaaS US High Nee SeMI Technologies Maker van Weaviate open-source vector database. https://www.weaviate.io Vector Database Open Source Netherlands Medium Nee Qdrant Open-source vector zoekmachine. https://www.qdrant.tech Vector Database Open Source Germany Medium Nee Chroma Open-source embedding database. https://www.trychroma.com Vector Database Open Source US Low Nee Neo4j Inc. Marktleider in graph databases. https://www.neo4j.com Knowledge Graph Commercial Sweden/US High Nee Ontotext Semantic graph database en kennistechnologie. https://www.ontotext.com Knowledge Graph Commercial Bulgaria Medium Nee Apache Foundation Open-source softwarestichting, o.a. Jena triple store. https://jena.apache.org Open Source Stichting Open Source US High Nee Stardog Enterprise knowledge graph platform. https://www.stardog.com Knowledge Graph Commercial US Medium Nee Stanford / W3C Academic ontology editor Protege. https://protege.stanford.edu Academisch Open Source US Medium Nee Triply Netherlandss bedrijf, linked data platform TriplyDB. https://www.triply.cc Knowledge Graph Commercial Netherlands Medium Nee Databricks Unified data analytics en AI-platform. https://www.databricks.com Data Platform Commercial SaaS US High Nee Snowflake Cloud data warehouse platform. https://www.snowflake.com Data Platform Commercial SaaS US High Nee MLflow (Linux Foundation) Open-source ML lifecycle management. https://www.mlflow.org MLOps Open Source US High Nee Weights & Biases ML experiment tracking en model management. https://www.wandb.ai MLOps Commercial SaaS US Medium Nee Kubeflow (Linux Foundation) ML pipeline orkestratie op Kubernetes. https://www.kubeflow.org MLOps Open Source US Medium Nee Neptune.ai Experiment tracking voor ML-teams. https://www.neptune.ai MLOps Commercial SaaS Poland Medium Nee ClearML Open-source MLOps platform. https://www.clear.ml MLOps Open Source Israel Medium Nee DVC (Iterative) Data en model version control. https://www.dvc.org MLOps Open Source US Medium Nee Hugging Face Open-source AI-community, model hub en tooling. https://www.huggingface.co AI Platform Open Source US/France High Yes Guardrails AI Output validation and filtering for LLMs. https://www.guardrailsai.com AI Safety Open Source US Low Yes Garak Project LLM vulnerability scanning toolkit. https://github.com/leondz/garak AI Safety Open Source UK Low Yes PromptFoo LLM evaluatie en red teaming tool. https://www.promptfoo.dev AI Evaluation Open Source US Low Nee RAGAS Project RAG-evaluatie framework. https://www.ragas.io AI Evaluation Open Source India Low Nee Confident AI LLM-evaluatieplatform DeepEval. https://www.confident-ai.com AI Evaluation Commercial SaaS US Low Nee TruEra AI quality en observability platform. https://www.truera.com AI Observability Commercial SaaS US Medium Nee Stanford CRFM Center for Research on Foundation Models. https://crfm.stanford.edu Academisch Open Source US Low Nee EleutherAI Open-source AI-onderzoekslaboratorium. https://www.eleuther.ai Research Open Source US Low Nee Prometheus (CNCF) Open-source monitoring en alerting. https://www.prometheus.io Monitoring Open Source US High Nee Grafana Labs Observability platform, dashboards en logging. https://www.grafana.com Monitoring Open Source Sweden High Nee Datadog Cloud monitoring en analytics platform. https://www.datadoghq.com Monitoring Commercial SaaS US High Nee Elastic Zoek- en observability platform (ELK Stack). https://www.elastic.co Monitoring Open Source Netherlands/US High Nee OpenTelemetry (CNCF) Open-source observability framework. https://www.opentelemetry.io Monitoring Open Source US High Nee Jaeger (CNCF) Distributed tracing systeem. https://www.jaegertracing.io Monitoring Open Source US Medium Nee Evidently AI ML monitoring en data drift detectie. https://www.evidentlyai.com AI Monitoring Open Source US Medium Yes WhyLabs AI observability en data monitoring. https://www.whylabs.ai AI Monitoring Commercial SaaS US Medium Yes NannyML Post-deployment ML performance monitoring. https://www.nannyml.com AI Monitoring Open Source Belgium Medium Yes Helicone LLM observability en token tracking. https://www.helicone.ai AI Monitoring Commercial SaaS US Low Nee LLMeter Project LLM inference cost benchmarking. https://github.com/aws-samples/llmeter AI Monitoring Open Source US Low Nee PromptLayer Prompt versiebeheer en evaluatie. https://www.promptlayer.com Prompt Management Commercial SaaS US Low Nee Humanloop LLM-applicatie ontwikkelplatform. https://www.humanloop.com Prompt Management Commercial SaaS UK Medium Nee Kong Inc. API gateway en service mesh. https://www.konghq.com API Gateway Open Source US High Nee LiteLLM Unified LLM API proxy en load balancer. https://www.litellm.ai API Gateway Open Source US Low Nee OpenRouter Unified API voor meerdere LLM-providers. https://www.openrouter.ai API Gateway Commercial SaaS US Low Nee Keycloak (Red Hat) Open-source identity en access management. https://www.keycloak.org Identity & Access Open Source US High Nee Okta Enterprise identity platform. https://www.okta.com Identity & Access Commercial SaaS US High Nee Open Policy Agent (CNCF) Unified policy engine, policy-as-code. https://www.openpolicyagent.org Policy Engine Open Source US High Nee Cedar (AWS) Policy language voor autorisatielogica. https://www.cedarpolicy.com Policy Engine Open Source US Medium Nee IBM AI Fairness 360 toolkit voor biasdetectie. https://aif360.mybluemix.net AI Platform Open Source US High Yes Aequitas Project Open-source bias audit toolkit. https://aequitas.dssg.io AI Fairness Open Source US Low Yes SHAP Project SHapley Additive exPlanations voor model interpretatie. https://shap.readthedocs.io Explainability Open Source US Medium Yes LIME Project Local Interpretable Model-agnostic Explanations. https://github.com/marcotcr/lime Explainability Open Source US Medium Yes Great Expectations Data validation and documentation framework. https://www.greatexpectations.io Data Quality Open Source US Medium Nee Soda Data quality monitoring platform. https://www.soda.io Data Quality Commercial SaaS Belgium Medium Nee HumanSignal Maker van Label Studio, open-source annotatie. https://www.humansignal.com Annotation Open Source US Medium Nee Explosion AI Creator of spaCy and Prodigy annotation tool. https://www.explosion.ai NLP Tooling Commercial Germany Medium Nee TensorFlow (Google) Open-source ML framework en serving. https://www.tensorflow.org ML Framework Open Source US High Nee PyTorch (Meta) Open-source deep learning framework. https://www.pytorch.org ML Framework Open Source US High Nee GitHub Code hosting en CI/CD platform. https://github.com DevOps Platform Commercial SaaS US High Nee GitLab DevOps platform met geïntegreerde CI/CD. https://www.gitlab.com DevOps Platform Open Source US High Nee Jenkins (Linux Foundation) Open-source automation server. https://www.jenkins.io CI/CD Open Source US High Nee Docker Inc. Containerisatieplatform. https://www.docker.com Container Platform Open Source US High Nee Kubernetes (CNCF) Container orkestratie standaard. https://www.kubernetes.io Container Orchestratie Open Source US High Nee Helm (CNCF) Kubernetes package manager. https://www.helm.sh Package Management Open Source US High Nee HashiCorp Infrastructure-as-code (Terraform). https://www.hashicorp.com Infrastructure Open Source US High Nee Pulumi Infrastructure-as-code in programmeertalen. https://www.pulumi.com Infrastructure Open Source US Medium Nee Atlassian Samenwerkingsplatform, maker van Confluence. https://www.atlassian.com Collaboration Commercial SaaS Australia High Nee Feast (Linux Foundation) Open-source feature store voor ML. https://www.feast.dev Feature Store Open Source US Medium Nee Tecton Enterprise feature platform voor ML. https://www.tecton.ai Feature Store Commercial SaaS US Medium Nee OpenLineage (Linux Foundation) Open-source data lineage standaard. https://www.openlineage.io Data Lineage Open Source US Medium Nee ArchiXL Dutch architecture firm, specialist in enterprise and information architecture. Creator of WikiXL, ArchiMedes and BegrippenXL. https://www.archixl.nl Architecture Commercial SaaS Netherlands Medium Nee Chief AI Officer (CAIO) Ultimately responsible for AI strategy, governance and value creation. Chair of the AI Board and guardian of AI-BOK implementation. Executive 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Portfolio Manager Manages the AI portfolio: selection, prioritisation and monitoring of AI initiatives based on business value and strategic fit. Senior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Ethics Advisor Advises on ethical aspects of AI systems: fairness, bias, transparency and societal impact. Member of the AI Ethics Committee. Senior 0,2-0,5 AI-BOK v1.2, Module 2, section 2.2 AI Product Owner Responsible for functional steering of an AI initiative: requirements, prioritisation and acceptance from the business perspective. Senior 1,0 AI-BOK v1.2, Module 2, section 2.2 AI Architect Designs the AI reference architecture: planes, ABBs, integration patterns and technology choices. Ensures coherence and reusability. Senior 1,0-2,0 AI-BOK v1.2, Module 2, section 2.2 AI Risk Manager Identifies, assesses and mitigates AI-related risks. Manages the AI risk register and advises on risk classification (EU AI Act). Senior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Governance Lead Operationalises AI governance: policy frameworks, mandates, quality gates and compliance processes. Member of the AI Review Committee. Medior 0,5 AI-BOK v1.2, Module 2, section 2.2 AI Knowledge Manager Curates and manages knowledge sources for AI grounding: terminology frameworks, knowledge graphs and document collections. Senior 1,0 AI-BOK v1.2, Module 2, section 2.2 AI Engineer Develops, trains and optimises AI models: data engineering, model training, fine-tuning, prompt engineering and evaluation. Medior 2,0-5,0 AI-BOK v1.2, Module 2, section 2.2 AI Interaction Designer Designs user interaction with AI systems: conversation flows, explainability, feedback mechanisms and human-in-the-loop patterns. Medior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Operations Engineer Responsible for deployment, monitoring and operational management of AI systems in production. Monitors SLAs and performance. Medior 1,0-3,0 AI-BOK v1.2, Module 2, section 2.2 AI Quality Analyst Evaluates and validates AI systems for quality: faithfulness, relevance, bias, robustness and conformity to quality gates. Medior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Auditor Performs independent audits on AI systems: compliance, technical conformity, data quality and governance adherence. Senior 0,2-0,5 AI-BOK v1.2, Module 2, section 2.2 AI System Administrator Manages the technical AI infrastructure: platforms, API gateways, model registries, monitoring tools and access control. Junior/Medior 1,0-2,0 AI-BOK v1.2, Module 2, section 2.2 AI Trainer/Annotator Prepares training data: labelling, annotation, quality control and validation of datasets for supervised learning. Junior 1,0-3,0 AI-BOK v1.2, Module 2, section 2.2 AI Domain Expert Provides domain knowledge for AI systems: output validation, curation of knowledge sources and assessment of domain-specific quality. Senior (domein) 0,2-0,5 AI-BOK v1.2, Module 2, section 2.2 AI Change Manager Guides organisational change in AI adoption: stakeholder management, training, communication and culture change. Senior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Agent Governor Manages autonomous AI agents: defines mandates, escalation rules, autonomy boundaries and monitors agent behaviour in production. Senior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 AI Compliance Officer Ensures compliance with AI legislation: EU AI Act, GDPR, sector-specific requirements and reporting obligations. Senior 0,5-1,0 AI-BOK v1.2, Module 2, section 2.2 Knowledge Engineer Designs and manages knowledge graphs and ontologies for AI grounding. AI-BOK v1.2, Module 2, section 2.2 RAG Architect Designs and optimises RAG pipelines and retrieval strategies. AI-BOK v1.2, Module 2, section 2.2 AI Fairness Engineer Specialist in bias-detectie, fairness-metrics en debiasing-technieken. AI-BOK v1.2, Module 2, section 2.2 Explainability Specialist Designt en implementeert uitlegbaarheidsmechanismen (SHAP, LIME). AI-BOK v1.2, Module 2, section 2.2 AI Security Architect Beveiligingsarchitectuur voor AI-systemen, prompt injection preventie. AI-BOK v1.2, Module 2, section 2.2 AI Platform Engineer Builds and manages the AI platform (MLOps, serving, monitoring). AI-BOK v1.2, Module 2, section 2.2 Content Curator Selects and qualifies knowledge sources for RAG pipelines. AI-BOK v1.2, Module 2, section 2.2 Semantic Specialist Expert in terminologieraamwerken, SKOS, linked data voor AI. AI-BOK v1.2, Module 2, section 2.2 AI Solution Architect Designt concrete AI-oplossingen binnen de referentiearchitectuur. AI-BOK v1.2, Module 2, section 2.2 Data Protection Officer Data Protection Officer, oversight of GDPR compliance in AI. AI-BOK v1.2, Module 2, section 2.2 AI Literacy Officer Accountable for AI Literacy & Capability Development (KA13). Maintains competency profiles, curriculum, literacy register, Article 4 evidence file, refresh adherence, and effectiveness measurement. Reports into the AI Board. Tactical/Cross-functional 20 AI-BOK v1.1 KA13 AI-BOK v1.1, Module 2, Role 20 Ideation & Exploration Identification of AI opportunities, feasibility analysis and alignment with business strategy. Deliverable: validated AI business case. Manual Human-in-command AI-BOK v1.2, Module 3, section 3.1.1 Design & Architecture Architecture design, model selection, data preparation and integration patterns. Deliverable: approved architecture blueprint. Partially automated Human-in-command AI-BOK v1.2, Module 3, section 3.1.2 Data Acquisition & Preparation Collection, cleaning, labelling and validation of training and evaluation data. Deliverable: qualified datasets. Largely automated Human-in-the-loop AI-BOK v1.2, Module 3, section 3.1.3 Model Development & Training Training, fine-tuning, prompt engineering and experimentation. Deliverable: trained and validated model with model card. Largely automated Human-in-the-loop AI-BOK v1.2, Module 3, section 3.1.4 Evaluation & Validation Quality evaluation on faithfulness, bias, robustness and conformity to quality gates. Deliverable: evaluation report and go/no-go. Automated with human review Human-in-the-loop AI-BOK v1.2, Module 3, section 3.1.5 Deployment & Integration Deployment to production, integration with business processes and user acceptance. Deliverable: operational AI system. Automated (CI/CD) Human-on-the-loop AI-BOK v1.2, Module 3, section 3.1.6 Monitoring & Operations Continuous monitoring of performance, drift, costs and anomalies. Triggers retraining loop on degradation. Deliverable: operational logs and alerts. Fully automated Human-on-the-loop AI-BOK v1.2, Module 3, section 3.1.7 Evolution & Decommissioning Evaluation of continued use, knowledge transfer and responsible decommissioning. Deliverable: lessons learned and decommissioning report. Manual met toolondersteuning Human-in-command AI-BOK v1.2, Module 3, section 3.1.8 TOGAF Integration AI Architecture as 5th layer (Cognition Plane). AI Architect on Enterprise Architecture Board. AI-BOK v1.2, Module 3, section 3.3 DAMA-DMBOK Integration KA5/KA12 extend data governance. Terminology frameworks in metadata management. AI-BOK v1.2, Module 3, section 3.3 COBIT/ITIL Integration AI systems in CMDB. AI incidents via existing incident management + AI classification. AI-BOK v1.2, Module 3, section 3.3 COSO/ISO 31000 Integration AI risk register integrated in enterprise risk register. Three Lines of Defence with AI-specific controls. AI-BOK v1.2, Module 3, section 3.3 GDPR Integration DPIA for AI systems. Art. 22 right to explanation operationalised. Privacy by design. AI-BOK v1.2, Module 3, section 3.3 Agile/SAFe Integration Gate criteria as Definition of Done per PI. MLOps pipeline as CD Pipeline equivalent. AI-BOK v1.2, Module 3, section 3.3 Retraining Loop P7 → P4: Performance degradation triggers retraining. AI-BOK v1.2, Module 3, section 3.1.9 Evaluation Loop P5 → P4: Gezakte evaluatie keert terug naar modelontwikkeling. AI-BOK v1.2, Module 3, section 3.1.9 Evolution Loop P8 → P1: Fundamental changes restart the full cycle. AI-BOK v1.2, Module 3, section 3.1.9 Data Bias Assessment Assessment per dataset on bias and representativeness. Possibly special category personal data Confidential AI-BOK v1.2, Module 1, section 5.8 Supply Chain Risk Assessment Risk assessment of external models, datasets, services. No personal data Confidential AI-BOK v1.2, Module 1, section 9.13 Gate: Business Case Approved AI-BOK v1.2, Module 3, section 3.1.1 Gate: Architecture Review AI-BOK v1.2, Module 3, section 3.1.2 Gate: Data Readiness AI-BOK v1.2, Module 3, section 3.1.3 Gate: Model Validated AI-BOK v1.2, Module 3, section 3.1.4 Gate: Production Ready AI-BOK v1.2, Module 3, section 3.1.5 Gate: Go Live AI-BOK v1.2, Module 3, section 3.1.6 Gate: Operations Stable AI-BOK v1.2, Module 3, section 3.1.7 Audit Trail Immutable log of AI decisions and interactions. Likely personal data (decision history) Confidential 10 years (EU AI Act) Mandatory (Art. 12) AI-BOK v1.2, Module 1, section 10.4 Dataset Verzameling data voor training, evaluatie of operationeel gebruik. AI-BOK v1.2, Module 3, section 3.2 Knowledge Source Externe of interne bron van gestructureerde kennis (ontologie, KG, vocabulaire). AI-BOK v1.2, Module 3, section 3.2 Prompt/Instruction Instructie of context die het gedrag van een AI-component stuurt. AI-BOK v1.2, Module 3, section 3.2 LLM/Foundation Model Service Inference endpoint for language models. Provides text, image and code generation via foundation models. Current Cognition Plane Commercial SaaS / On-premise REST API / OpenAI-compatible 99,9% (business hours) High (prompts may contain personal data) AI-BOK v1.2, Module 3, section 3.6.5a Agent Orchestrator Orchestration of autonomous AI agents, task decomposition, tool use and multi-agent coordination. Current Cognition Plane Open Source On-premise MCP / A2A 99,9% High (agent has access to business data) AI-BOK v1.2, Module 3, section 3.6.5a RAG Pipeline Retrieval-Augmented Generation: retrieves relevant context from knowledge sources and feeds it to the LLM. Current Cognition Plane Open Source On-premise REST API / Python SDK 99,9% High (knowledge sources may be confidential) AI-BOK v1.2, Module 3, section 3.6.5a Prompt Registry Central management of prompt templates, system messages, version control and effectiveness metrics. Current Cognition Plane Commercial SaaS / On-premise REST API AI-BOK v1.2, Module 3, section 3.6.5a Guardrails Engine Input/output filtering, content moderation, prompt injection detection and safety controls. Current Cognition Plane Open Source On-premise Python SDK / Middleware 99,99% (kritiek) AI-BOK v1.2, Module 3, section 3.6.5a Evaluation Framework Automated quality evaluation of models: faithfulness, relevance, hallucination rate. Current Cognition Plane Open Source On-premise Python SDK AI-BOK v1.2, Module 3, section 3.6.5a AI Identity & Access Mgmt Identity and authorisation for AI agents, users and systems. Agent identities (NHI). Current Control Plane Commercial SaaS OAuth 2.0 / OIDC / SAML 99,99% (kritiek) Very high (identity data) AI-BOK v1.2, Module 3, section 3.6.5a Policy Engine Policy rules for AI authorisation, policy-as-code. Runtime governance checks for agents. Current Control Plane Open Source On-premise REST API / Rego / Cedar 99,99% (kritiek) AI-BOK v1.2, Module 3, section 3.6.5a Audit Trail Service Immutable logging of AI decisions, interactions and agent actions. Provenance and traceability. Current Control Plane Open Source On-premise OpenTelemetry / REST API 99,999% (compliance) High (contains decision history) AI-BOK v1.2, Module 3, section 3.6.5a Model Registry Central registry of models, versions, model cards, experiments and deployments. Current Control Plane Open Source On-premise / SaaS REST API / MLflow API 99,9% AI-BOK v1.2, Module 3, section 3.6.5a API Gateway Access control, rate limiting, routing and load balancing for AI services and LLM endpoints. Current Control Plane Commercial On-premise / SaaS REST / gRPC / GraphQL 99,99% AI-BOK v1.2, Module 3, section 3.6.5a Vector Database Storage and similarity search of embeddings for RAG and semantic search. Current Data Plane Commercial SaaS / On-premise REST API / gRPC 99,9% High (embeddings of business data) AI-BOK v1.2, Module 3, section 3.6.5a Knowledge Graph Knowledge graph for structured knowledge: entities, relations, ontologies (RDF/OWL/SKOS). Current Data Plane Open Source On-premise SPARQL / REST API 99,9% Medium (concepts and relationships) AI-BOK v1.2, Module 3, section 3.6.5a Feature Store Standardised storage and serving of ML features for training and inference. Current Data Plane Open Source On-premise REST API / gRPC AI-BOK v1.2, Module 3, section 3.6.5a Data Lake/Warehouse Central storage for training data, evaluation data and analytics. Current Data Plane Commercial SaaS SQL / REST API 99,9% Very high (training data, personal data) AI-BOK v1.2, Module 3, section 3.6.5a Document Store Document storage for RAG sources: policy documents, manuals, knowledge articles. Current Data Plane Commercial SaaS / On-premise REST API / WebDAV High (policy documents, knowledge sources) AI-BOK v1.2, Module 3, section 3.6.5a Monitoring & Observability Continuous monitoring of AI systems: performance, costs, drift, anomalies and SLA monitoring. Current Control Plane Open Source SaaS / On-premise OpenTelemetry / Prometheus 99,99% AI-BOK v1.2, Module 3, section 3.6.5a Bias Detection & Fairness Detection and mitigation of bias in models and data. Fairness metrics and debiasing techniques. Current Cognition Plane Open Source On-premise AI-BOK v1.2, Module 3, section 3.6.5a Explainability Engine Uitlegbaarheidsmechanismen voor AI-beslissingen: feature attribution, counterfactuals, attention maps. Current Cognition Plane Open Source On-premise AI-BOK v1.2, Module 3, section 3.6.5a Data Quality Framework Geautomatiseerde datakwaliteitsValidation: volledigheid, consistentie, actualiteit, biasdetectie. Current Data Plane Open Source On-premise AI-BOK v1.2, Module 3, section 3.6.5a Annotation Tooling Tooling for data annotation, labelling and curation. Support for RLHF and domain expert validation. Current Data Plane Open Source On-premise / SaaS AI-BOK v1.2, Module 3, section 3.6.5a CI/CD & Infrastructure Continuous integration/delivery pipelines, containerisatie, infrastructure-as-code voor AI-workloads. Current Control Plane Open Source SaaS / On-premise YAML / REST API AI-BOK v1.2, Module 3, section 3.6.5a Model Serving Production-serving van ML-modellen: batching, caching, GPU-optimalisatie, A/B-testing. Current Control Plane Open Source On-premise AI-BOK v1.2, Module 3, section 3.6.5a Naive RAG Basis retrieve-then-generate patroon. AI-BOK v1.2, Module 1, section 3.6 Advanced RAG Pre-retrieval optimalisatie, post-retrieval reranking. AI-BOK v1.2, Module 1, section 3.6 Modular RAG Modulaire pipeline met verwisselbare componenten. AI-BOK v1.2, Module 1, section 3.6 GraphRAG Knowledge Graph-gebaseerde retrieval met entity-relatie context. AI-BOK v1.2, Module 1, section 3.6 Agentic RAG Agent-driven retrieval with tool use and iterative search strategies. AI-BOK v1.2, Module 1, section 3.6 AI System Een gedeployd AI-systeem dat als geheel functioneert. AI-BOK v1.2, Module 3, section 3.2 Agent Autonome AI-component die zelfstandig redeneert en handelt binnen mandaat. AI-BOK v1.2, Module 3, section 3.2 AI Model Getraind ML/AI-model in deployable vorm. AI-BOK v1.2, Module 3, section 3.2 Data Retrieval Service Vector similarity search and document retrieval for RAG pipelines. Data Plane AI-BOK v1.2, Module 3, section 3.6.5 Knowledge Query Service SPARQL and graph queries on knowledge graphs and ontologies. Data Plane AI-BOK v1.2, Module 3, section 3.6.5 Data Quality Service Geautomatiseerde datakwaliteitsValidation en -rapportage. Data Plane AI-BOK v1.2, Module 3, section 3.6.5 Feature Serving Service On-demand serving of ML features for training and inference. Data Plane AI-BOK v1.2, Module 3, section 3.6.5 Identity & Authorization Service Agent-identiteit, autorisatie en mandaatverificatie. Control Plane AI-BOK v1.2, Module 3, section 3.6.5 Policy Check Service Runtime governance checks: may this agent perform this action? Control Plane AI-BOK v1.2, Module 3, section 3.6.5 Audit & Logging Service Immutable logging of decisions, interactions and agent actions. Control Plane AI-BOK v1.2, Module 3, section 3.6.5 Model Management Service Modelselectie, versiebeheer en model card raadpleging. Control Plane AI-BOK v1.2, Module 3, section 3.6.5 API Access Service Controlled access to AI endpoints with rate limiting and routing. Control Plane AI-BOK v1.2, Module 3, section 3.6.5 Monitoring & Observability Service Continuous monitoring of performance, drift, costs and anomalies. Control Plane AI-BOK v1.2, Module 3, section 3.6.5 AI Inference Service Tekstgeneratie, classificatie, embedding via foundation models. Cognition Plane AI-BOK v1.2, Module 3, section 3.6.5 Agent Orchestration Service Multi-agent coördinatie, taakdecompositie en tool-gebruik. Cognition Plane AI-BOK v1.2, Module 3, section 3.6.5 Knowledge Retrieval Service RAG: grounded answers based on authoritative knowledge sources. Cognition Plane AI-BOK v1.2, Module 3, section 3.6.5 Guardrails & Safety Service Input/output-filtering, Hallucinationpreventie en content moderation. Cognition Plane AI-BOK v1.2, Module 3, section 3.6.5 AI Evaluation Service Geautomatiseerde kwaliteitsevaluatie: faithfulness, relevance, fairness. Cognition Plane AI-BOK v1.2, Module 3, section 3.6.5 Pseudonymisation Service De-pseudonymisation only possible by authorised roles with specific task key. AI-BOK v1.2, Module 1, section 5.10 Key Vault Pattern Pseudonymisation layer separating personal data from content data. Task keys link pseudonymised data to tasks, not to persons. AI-BOK v1.2, Module 1, section 5.10 OpenAI GPT-4o Current High AI-BOK v1.2, Module 3, section 3.6.5a Anthropic Claude Current High AI-BOK v1.2, Module 3, section 3.6.5a Meta LLaMA Current Free AI-BOK v1.2, Module 3, section 3.6.5a Google Gemini Current High AI-BOK v1.2, Module 3, section 3.6.5a Cohere Medium AI-BOK v1.2, Module 3, section 3.6.5a Azure OpenAI Service High AI-BOK v1.2, Module 3, section 3.6.5a AWS Bedrock High AI-BOK v1.2, Module 3, section 3.6.5a vLLM Planned Free AI-BOK v1.2, Module 3, section 3.6.5a TensorFlow Serving AI-BOK v1.2, Module 3, section 3.6.5a TorchServe AI-BOK v1.2, Module 3, section 3.6.5a Triton Inference Server High AI-BOK v1.2, Module 3, section 3.6.5a LangGraph Current AI-BOK v1.2, Module 3, section 3.6.5a CrewAI Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a AutoGen Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a Semantic Kernel Emerging AI-BOK v1.2, Module 3, section 3.6.5a LangChain Current Free AI-BOK v1.2, Module 3, section 3.6.5a LlamaIndex Current Free AI-BOK v1.2, Module 3, section 3.6.5a Haystack Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a Azure AI Search High AI-BOK v1.2, Module 3, section 3.6.5a Amazon Kendra High AI-BOK v1.2, Module 3, section 3.6.5a PromptLayer Low AI-BOK v1.2, Module 3, section 3.6.5a Humanloop Low AI-BOK v1.2, Module 3, section 3.6.5a NeMo Guardrails Current Free AI-BOK v1.2, Module 3, section 3.6.5a Guardrails AI Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a Garak Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a PromptFoo Emerging Medium AI-BOK v1.2, Module 3, section 3.6.5a RAGAS Current Free AI-BOK v1.2, Module 3, section 3.6.5a DeepEval Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a TruLens Emerging AI-BOK v1.2, Module 3, section 3.6.5a HELM AI-BOK v1.2, Module 3, section 3.6.5a lm-eval-harness AI-BOK v1.2, Module 3, section 3.6.5a BLEU AI-BOK v1.2, Module 3, section 3.6.5a ROUGE AI-BOK v1.2, Module 3, section 3.6.5a BERTScore AI-BOK v1.2, Module 3, section 3.6.5a Microsoft Entra ID Current High AI-BOK v1.2, Module 3, section 3.6.5a Keycloak Free AI-BOK v1.2, Module 3, section 3.6.5a Okta Medium AI-BOK v1.2, Module 3, section 3.6.5a Open Policy Agent Current Free AI-BOK v1.2, Module 3, section 3.6.5a Cedar Planned AI-BOK v1.2, Module 3, section 3.6.5a OpenTelemetry Current Free AI-BOK v1.2, Module 3, section 3.6.5a Jaeger Free AI-BOK v1.2, Module 3, section 3.6.5a Langfuse Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a OpenLineage AI-BOK v1.2, Module 3, section 3.6.5a MLflow Current Medium AI-BOK v1.2, Module 3, section 3.6.5a Weights & Biases Medium AI-BOK v1.2, Module 3, section 3.6.5a Kubeflow AI-BOK v1.2, Module 3, section 3.6.5a Google Vertex AI High AI-BOK v1.2, Module 3, section 3.6.5a Azure ML Model Registry AI-BOK v1.2, Module 3, section 3.6.5a Neptune.ai Medium AI-BOK v1.2, Module 3, section 3.6.5a ClearML Medium AI-BOK v1.2, Module 3, section 3.6.5a DVC (Data Version Control) Free AI-BOK v1.2, Module 3, section 3.6.5a Hugging Face Model Hub AI-BOK v1.2, Module 3, section 3.6.5a Model Card Toolkit AI-BOK v1.2, Module 3, section 3.6.5a Kong Medium AI-BOK v1.2, Module 3, section 3.6.5a Apigee High AI-BOK v1.2, Module 3, section 3.6.5a AWS API Gateway AI-BOK v1.2, Module 3, section 3.6.5a Azure API Management AI-BOK v1.2, Module 3, section 3.6.5a LiteLLM Emerging Low AI-BOK v1.2, Module 3, section 3.6.5a OpenRouter Emerging Low AI-BOK v1.2, Module 3, section 3.6.5a Pinecone Current Medium AI-BOK v1.2, Module 3, section 3.6.5a Weaviate Current Low AI-BOK v1.2, Module 3, section 3.6.5a Qdrant Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a ChromaDB Emerging Free AI-BOK v1.2, Module 3, section 3.6.5a pgvector Free AI-BOK v1.2, Module 3, section 3.6.5a Neo4j Current Low AI-BOK v1.2, Module 3, section 3.6.5a Amazon Neptune AI-BOK v1.2, Module 3, section 3.6.5a GraphDB Low AI-BOK v1.2, Module 3, section 3.6.5a Apache Jena Free AI-BOK v1.2, Module 3, section 3.6.5a Stardog Low AI-BOK v1.2, Module 3, section 3.6.5a Protégé Free AI-BOK v1.2, Module 3, section 3.6.5a Databricks Current High AI-BOK v1.2, Module 3, section 3.6.5a Snowflake High AI-BOK v1.2, Module 3, section 3.6.5a Delta Lake AI-BOK v1.2, Module 3, section 3.6.5a Feast Planned Free AI-BOK v1.2, Module 3, section 3.6.5a Tecton Planned AI-BOK v1.2, Module 3, section 3.6.5a SharePoint AI-BOK v1.2, Module 3, section 3.6.5a Confluence AI-BOK v1.2, Module 3, section 3.6.5a Amazon S3 AI-BOK v1.2, Module 3, section 3.6.5a Prometheus Current Free AI-BOK v1.2, Module 3, section 3.6.5a Grafana Current Free AI-BOK v1.2, Module 3, section 3.6.5a Datadog High AI-BOK v1.2, Module 3, section 3.6.5a Azure Monitor AI-BOK v1.2, Module 3, section 3.6.5a ELK Stack AI-BOK v1.2, Module 3, section 3.6.5a Loki Free AI-BOK v1.2, Module 3, section 3.6.5a Evidently AI Emerging Medium AI-BOK v1.2, Module 3, section 3.6.5a WhyLabs Emerging Medium AI-BOK v1.2, Module 3, section 3.6.5a NannyML Emerging Medium AI-BOK v1.2, Module 3, section 3.6.5a Helicone AI-BOK v1.2, Module 3, section 3.6.5a LLMeter AI-BOK v1.2, Module 3, section 3.6.5a Azure Cost Management AI-BOK v1.2, Module 3, section 3.6.5a AWS Cost Explorer AI-BOK v1.2, Module 3, section 3.6.5a AI Fairness 360 (IBM) Free AI-BOK v1.2, Module 3, section 3.6.5a Fairlearn (Microsoft) Free AI-BOK v1.2, Module 3, section 3.6.5a Aequitas Free AI-BOK v1.2, Module 3, section 3.6.5a What-If Tool (Google) AI-BOK v1.2, Module 3, section 3.6.5a SHAP Free AI-BOK v1.2, Module 3, section 3.6.5a LIME Free AI-BOK v1.2, Module 3, section 3.6.5a Great Expectations Free AI-BOK v1.2, Module 3, section 3.6.5a Soda Free AI-BOK v1.2, Module 3, section 3.6.5a Label Studio Free AI-BOK v1.2, Module 3, section 3.6.5a Prodigy AI-BOK v1.2, Module 3, section 3.6.5a GitHub Actions Free AI-BOK v1.2, Module 3, section 3.6.5a GitLab CI AI-BOK v1.2, Module 3, section 3.6.5a Azure DevOps AI-BOK v1.2, Module 3, section 3.6.5a Jenkins Free AI-BOK v1.2, Module 3, section 3.6.5a Docker Current Free AI-BOK v1.2, Module 3, section 3.6.5a Kubernetes Current Free AI-BOK v1.2, Module 3, section 3.6.5a Helm Charts Free AI-BOK v1.2, Module 3, section 3.6.5a Terraform Free AI-BOK v1.2, Module 3, section 3.6.5a Pulumi Planned AI-BOK v1.2, Module 3, section 3.6.5a TriplyDB Current Low AI-BOK v1.2, Module 3, section 3.6.5a BegrippenXL Current Free AI-BOK v1.2, Module 3, section 3.6.5a WikiXL Current Low AI-BOK v1.2, Module 3, section 3.6.5a ArchiMedes Current Low AI-BOK v1.2, Module 3, section 3.6.5a EU AI Act European regulation for AI with risk classes. AI-BOK v1.2, Module 1, section 1.1 Digital Transformation Organisations must leverage AI to remain competitive. AI-BOK v1.2, Module 1, section 2.1 Trust in AI Public trust in AI is essential for adoption. AI-BOK v1.2, Module 1, section 11.1 Rise of Agentic AI Autonomous AI agents require new governance mechanisms. AI-BOK v1.2, Module 3, section 3.6.6 Technological Acceleration Rapid development of foundation models and agentic AI. AI-BOK v1.2, Module 3, section 3.4.8 Talent Scarcity Labour market for AI specialists is very tight. AI-BOK v1.2, Module 3, section 3.4.8 Organisational Culture & Change Readiness Culture determines AI adoption pace. AI-BOK v1.2, Module 3, section 3.4.8 Sector-specific Regulation Additional requirements per sector (healthcare, finance, government). AI-BOK v1.2, Module 3, section 3.4.8 Sustainability & ESG (CSRD) AI energy consumption and CO2 emissions are reported. AI-BOK v1.2, Module 3, section 3.4.8 Data Spaces & Supply Chain Digitisation Federated data sharing requires interoperable AI. AI-BOK v1.2, Module 3, section 3.4.8 Governance Framework Het geheel van beleid, principes en mandaten dat AI-inzet stuurt. AI-BOK v1.2, Module 3, section 3.2 KA1: Responsible AI Governance Create conditions under which AI systems deliver maximum value within acceptable risk boundaries. AI-BOK v1.2, Module 1, section 1.2 KA2: Maximum AI Value Creation Maximum organisational value from AI investments through systematic selection and prioritisation. AI-BOK v1.2, Module 1, section 2.2 KA3: Coherent AI Architecture A coherent, repeatable and evolvable blueprint for AI capabilities. AI-BOK v1.2, Module 1, section 3.2 KA4: Predictable AI Lifecycle Predictable, repeatable and governable progress of AI initiatives. AI-BOK v1.2, Module 1, section 4.2 KA5: Reliable Data Foundation Reliable, traceable and semantically anchored data foundation for AI. AI-BOK v1.2, Module 1, section 5.2 KA6: Manageable Model Portfolio Reliable, reproducible and responsible model portfolio. AI-BOK v1.2, Module 1, section 6.2 KA7: User-Centred AI Interaction AI interactions that are user-centred, reliable and inclusive. AI-BOK v1.2, Module 1, section 7.2 KA8: Reliable AI Operations Reliable, observable and cost-efficient AI operations in production. AI-BOK v1.2, Module 1, section 8.2 KA9: Proactive Risk Management Proactively identify, assess and mitigate AI-related risks. AI-BOK v1.2, Module 1, section 9.2 KA10: Demonstrable AI Compliance Verifiable, risk-based compliance of AI systems. AI-BOK v1.2, Module 1, section 10.2 KA11: Ethically Responsible AI Ethical considerations structurally embedded in the design and deployment of AI. AI-BOK v1.2, Module 1, section 11.2 KA12: Grounded AI Knowledge Processing AI systems produce grounded, traceable and semantically vamemberated output. AI-BOK v1.2, Module 1, section 12.2 Responsible AI Deployment AI is deployed in a responsible, transparent and safe manner. AI-BOK v1.2, Module 3, section 3.0 AI Compliance Full compliance with EU AI Act and relevant standards. AI-BOK v1.2, Module 3, section 3.0 KA13: AI-literate organisation Semantic Precision Unambiguous definitions of the concepts AI uses. AI-BOK v1.2, Module 3, section 3.6.4 Reliable Knowledge Sources Authoritative, curated information for grounding. AI-BOK v1.2, Module 3, section 3.6.4 Explicit Governance Clear rules, roles and mandates. AI-BOK v1.2, Module 3, section 3.6.4 Responsibility is explicitly assigned Every AI system has a designated owner responsible for functioning, impact and policy compliance. AI-BOK v1.2, Module 1, section 1.2 Governance is proportional to risk Governance intensity is proportional to the risk classification of the AI system. AI-BOK v1.2, Module 1, section 1.2 Transparency about autonomy The degree of autonomous operation is documented and visible. AI-BOK v1.2, Module 1, section 1.2 Governance is embedded, not imposed Governance mechanisms are part of the AI architecture, not an afterthought. AI-BOK v1.2, Module 1, section 1.2 Human control as a design principle Human-in-the-loop or human-on-the-loop is built in for decisions with significant impact. AI-BOK v1.2, Module 1, section 1.2 Continuous improvement based on metrics Governance is not static but is periodically adapted. AI-BOK v1.2, Module 1, section 1.2 Strategic alignment over technological ambition AI initiatives are selected based on contribution to business objectives. AI-BOK v1.2, Module 1, section 2.2 Portfolio balance as a steering mechanism The AI portfolio is deliberately balanced between explorative and exploitative. AI-BOK v1.2, Module 1, section 2.2 Value realisation over activity What counts is not the number of AI projects but the realised value. AI-BOK v1.2, Module 1, section 2.2 Capacity planning as a precondition AI strategy without capacity planning is a wish list. AI-BOK v1.2, Module 1, section 2.2 Adaptive planning in a dynamic field Strategy is reviewed annually; portfolio at least quarterly. AI-BOK v1.2, Module 1, section 2.2 Cognition plane as a strategic layer Strategy addresses the cognition plane as a new architecture layer. AI-BOK v1.2, Module 1, section 2.2 Semantic precision over ambiguity AI-systemen moeten concepten eenduidig interpreteren via terminologieraamwerken. AI-BOK v1.2, Module 1, section 3.2 Explicit governance over implicit control Authorities and escalation rules for AI agents are explicit in the architecture. AI-BOK v1.2, Module 1, section 3.2 Separation of responsibilities Architecture separates data, knowledge, logic and governance as four layers. AI-BOK v1.2, Module 1, section 3.2 Traceability as an architecture requirement Every AI output must be traceable to sources, models and reasoning steps. AI-BOK v1.2, Module 1, section 3.2 Interoperability through open standards AI components communicate via MCP and A2A. Vendor lock-in is minimised. AI-BOK v1.2, Module 1, section 3.2 Design for evolution Architecture is modular and loosely coupled so that components are replaceable. AI-BOK v1.2, Module 1, section 3.2 Privacy by design through pseudonymisation Personal data is never unnecessarily exposed to AI components. AI-BOK v1.2, Module 1, section 3.2 Phase-aware governance Each lifecycle phase has its own quality criteria, roles and deliverables. AI-BOK v1.2, Module 1, section 4.2 Continuous feedback loops Production observations structurally flow back to earlier phases. AI-BOK v1.2, Module 1, section 4.2 Reproducibility as a precondition Every experiment, training and deployment must be repeatable. AI-BOK v1.2, Module 1, section 4.2 Value-driven prioritisation The lifecycle is driven by business value, not just technical feasibility. AI-BOK v1.2, Module 1, section 4.2 Responsible decommissioning End of lifecycle is as important as the beginning: audit trail, data retention, knowledge transfer. AI-BOK v1.2, Module 1, section 4.2 Agentic awareness For agentic AI, additional lifecycle requirements apply per phase. AI-BOK v1.2, Module 1, section 4.2 Semantic precision over data volume Data value for AI is determined by unambiguity, not by volume. AI-BOK v1.2, Module 1, section 5.2 Grounding in authoritative sources AI-systemen verankeren antwoorden in traceerbare, autoritatieve kennisbronnen. AI-BOK v1.2, Module 1, section 5.2 Traceability from source to output Every AI output is traceable to data sources and semantic definitions. AI-BOK v1.2, Module 1, section 5.2 Data quality as a continuous process Data quality is an ongoing process of measurement, detection and improvement. AI-BOK v1.2, Module 1, section 5.2 Bias awareness in all data phases Bias detection is built into every phase of the data pipeline. AI-BOK v1.2, Module 1, section 5.2 Interoperability through open standards Semantische modellen gebruiken SKOS, RDF, OWL, SHACL. AI-BOK v1.2, Module 1, section 5.2 Every model is a managed asset Every production model is registered in a central model registry. AI-BOK v1.2, Module 1, section 6.2 Documentation as a first requirement No model to production without a model card. AI-BOK v1.2, Module 1, section 6.2 Reproducibility is non-negotiable Every model version must be reproducible. AI-BOK v1.2, Module 1, section 6.2 Bias testing is structural Modellen worden periodiek getest op ongewenste biases. AI-BOK v1.2, Module 1, section 6.2 Deliberate choice: foundation model vs custom model The choice is an architecture decision with consequences. AI-BOK v1.2, Module 1, section 6.2 Model portfolio thinking Models are managed as a portfolio: overlap and risks are visible. AI-BOK v1.2, Module 1, section 6.2 User first Every AI interaction is designed from the perspective of task, context and knowledge level. AI-BOK v1.2, Module 1, section 7.2 Trust calibration Users form a realistic picture of what AI can and cannot do. AI-BOK v1.2, Module 1, section 7.2 Transparency in operation The system proactively communicates about sources and uncertainties. AI-BOK v1.2, Module 1, section 7.2 Inclusive accessibility AI interfaces are usable for users with diverse abilities. AI-BOK v1.2, Module 1, section 7.2 Human autonomy The user retains control over the decision-making process. AI-BOK v1.2, Module 1, section 7.2 Iterative improvement Interactionpatronen worden continu geëvalueerd en verbeterd. AI-BOK v1.2, Module 1, section 7.2 Production reliability AI systems function with the same reliability as business-critical applications. AI-BOK v1.2, Module 1, section 8.2 Observability Performance, costs, quality and deviations are continuously measured. AI-BOK v1.2, Module 1, section 8.2 Reproducibility Every deployment is reproducible: models, configurations and prompts are versioned. AI-BOK v1.2, Module 1, section 8.2 Automated delivery CI/CD pipelines for maximum automation from model to production. AI-BOK v1.2, Module 1, section 8.2 Cost management AI-inferentiekosten worden actief gemonitord en geoptimaliseerd. AI-BOK v1.2, Module 1, section 8.2 Gradual rollout New versions are rolled out in phases (canary, A/B, blue-green). AI-BOK v1.2, Module 1, section 8.2 Risk proportionality Risk management intensity is proportional to the risk level. AI-BOK v1.2, Module 1, section 9.2 Proactive identification Risico's worden vóór deployment geïdentificeerd en geadresseerd. AI-BOK v1.2, Module 1, section 9.2 Continuous vigilance Risk profiles change; monitoring is continuous, not one-off. AI-BOK v1.2, Module 1, section 9.2 Safety as culture AI safety is an organisational culture, not a checklist. AI-BOK v1.2, Module 1, section 9.2 Legal conformity EU AI Act risk classification is the minimum framework. AI-BOK v1.2, Module 1, section 9.2 Transparent accountability Risk assessments are documented and available for audit. AI-BOK v1.2, Module 1, section 9.2 Demonstrable compliance Compliance is only real when verifiable and reproducible. AI-BOK v1.2, Module 1, section 10.2 Risk-based classification Compliance effort scales with risk level. AI-BOK v1.2, Module 1, section 10.2 Independence of audit Audits worden uitgevoerd door functioneel onafhankelijke partijen. AI-BOK v1.2, Module 1, section 10.2 Continuous traceability Audittrails worden continu vastgelegd, niet achteraf gereconstrueerd. AI-BOK v1.2, Module 1, section 10.2 Proactive transparency Organisaties publiceren actief in het Algorithmregister. AI-BOK v1.2, Module 1, section 10.2 Integral alignment AI Compliance integreert met enterprise governance en informatiebeveiliging. AI-BOK v1.2, Module 1, section 10.2 Fairness as a measurable requirement Fairness is a testable property with standardised metrics. AI-BOK v1.2, Module 1, section 11.2 Explainability as a right Stakeholders have the right to an understandable explanation of AI decisions. AI-BOK v1.2, Module 1, section 11.2 Human oversight as a fundamental principle Human-in/on/in-command per AI system based on risk and impact. AI-BOK v1.2, Module 1, section 11.2 Societal responsibility Organisaties beoordelen ook bredere maatschappelijke consequenties. AI-BOK v1.2, Module 1, section 11.2 Value alignment as a design requirement AI systems are designed in line with organisational and societal values. AI-BOK v1.2, Module 1, section 11.2 Ethics by design Ethical considerations are structurally integrated into the design process. AI-BOK v1.2, Module 1, section 11.2 Grounding over generation AI output based on verified, authoritative sources. AI-BOK v1.2, Module 1, section 12.2 Explicit source authority Knowledge sources are classified by reliability and authority. AI-BOK v1.2, Module 1, section 12.2 Hallucination prevention as a system requirement Prevention of factually incorrect output is a hard system requirement. AI-BOK v1.2, Module 1, section 12.2 Knowledge currency as a continuous process Mechanisms for staleness detection are structurally embedded. AI-BOK v1.2, Module 1, section 12.2 Semantic validation Every knowledge source and AI output is validated against terminology frameworks. AI-BOK v1.2, Module 1, section 12.2 Federated knowledge sharing Knowledge is unlocked at the source and shared according to data spaces principles. AI-BOK v1.2, Module 1, section 12.2 SKOS W3C standard for terminology frameworks and thesauri. AI-BOK v1.2, Module 3, section 3.3 NL-SBB (Geonovum) Standard for Describing Concepts. AI-BOK v1.2, Module 3, section 3.3 RDF / RDFS W3C Resource Description Framework for linked data. AI-BOK v1.2, Module 3, section 3.3 OWL W3C Web Ontology Language for ontologies. AI-BOK v1.2, Module 3, section 3.3 SHACL W3C Shapes Constraint Language for validation. AI-BOK v1.2, Module 3, section 3.3 SPARQL W3C query language for RDF data. AI-BOK v1.2, Module 3, section 3.3 DCAT 3.0 W3C Data Catalog Vocabulary for dataset metadata. AI-BOK v1.2, Module 3, section 3.3 DCAT-AP-NL Dutch profile for data catalogue. AI-BOK v1.2, Module 3, section 3.3 JSON-LD Linked data serialisation format. AI-BOK v1.2, Module 3, section 3.3 W3C PROV-O Provenance ontology for origin registration. AI-BOK v1.2, Module 3, section 3.3 Dublin Core Metadata standard for information sources. AI-BOK v1.2, Module 3, section 3.3 ISO/IEC 42001:2023 AI Management System standard. AI-BOK v1.2, Module 3, section 3.3 NIST AI RMF 1.0 AI Risk Management Framework. AI-BOK v1.2, Module 3, section 3.3 EU AI Act European regulation for AI governance. AI-BOK v1.2, Module 3, section 3.3 AVG / GDPR European privacy regulation. AI-BOK v1.2, Module 3, section 3.3 ISO/IEC 25010 Software product quality characteristics. AI-BOK v1.2, Module 3, section 3.3 IEEE P2247 Adaptive Autonomous Systems Architecture. AI-BOK v1.2, Module 3, section 3.3 TOGAF Enterprise Architecture framework. AI-BOK v1.2, Module 3, section 3.3 ArchiMate 3.2 Architecture modelling language. AI-BOK v1.2, Module 3, section 3.3 MIM Information Modelling Metamodel (Geonovum/VNG). AI-BOK v1.2, Module 3, section 3.3 OpenLineage Pipeline metadata lineage standard. AI-BOK v1.2, Module 3, section 3.3 MCP (Model Context Protocol) Agent-tool integration protocol. AI-BOK v1.2, Module 3, section 3.3 A2A (Agent-to-Agent) Agent-agent communication protocol. AI-BOK v1.2, Module 3, section 3.3 Model Cards (Mitchell et al.) De facto standard for model documentation. AI-BOK v1.2, Module 3, section 3.3 Datasheets for Datasets (Gebru) Standard for dataset documentation. AI-BOK v1.2, Module 3, section 3.3 Agent Card (A2A) Agent metadata specification. AI-BOK v1.2, Module 3, section 3.3 SKILL.md (Agent Skills) Open standard for AI interaction patterns. AI-BOK v1.2, Module 3, section 3.3 ISO/IEC 5338:2023 AI Lifecycle Processes. AI-BOK v1.2, Module 3, section 3.3 ISO/IEC 22989:2022 AI Concepts & Terminology. AI-BOK v1.2, Module 3, section 3.3 ISO/IEC 23894:2023 AI Risk Management. AI-BOK v1.2, Module 3, section 3.3 ISO/IEC 25012 Data Quality Model. AI-BOK v1.2, Module 3, section 3.3 ISO 9241-210:2019 Human-Centred Design. AI-BOK v1.2, Module 3, section 3.3 WCAG 2.2 / EN 301 549 Accessibility standard. AI-BOK v1.2, Module 3, section 3.3 FAIR Data Principles Findable, Accessible, Interoperable, Reusable. AI-BOK v1.2, Module 3, section 3.3 DMN (Decision Model & Notation) Standard for decision models. AI-BOK v1.2, Module 3, section 3.3 SBVR Semantics of Business Vocabulary and Business Rules. AI-BOK v1.2, Module 3, section 3.3 BPMN 2.0 Business Process Model and Notation. AI-BOK v1.2, Module 3, section 3.3 ONNX Open Neural Network Exchange format. AI-BOK v1.2, Module 3, section 3.3 OpenAPI Specification REST API specification standard. AI-BOK v1.2, Module 3, section 3.3 FinOps Framework Cloud financial management framework. AI-BOK v1.2, Module 3, section 3.3 ISO 31000:2018 Risk Management standard. AI-BOK v1.2, Module 3, section 3.3 EU AI Act Article 4 (AI literacy) Obligation for providers and deployers to ensure sufficient AI literacy of staff and other persons acting on their behalf. Effective 2 February 2025. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Digital Services Act (Regulation 2022/2065) Risk assessment, algorithmic transparency and recommender-system duties for VLOPs/VLOSEs. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Digital Markets Act (Regulation 2022/1925) Gatekeeper obligations including interoperability and self-preferencing prohibitions. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Data Act (Regulation 2023/2854) Rules on data access, sharing and switching that affect datasets feeding AI. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix NIS2 (Directive 2022/2555) Cybersecurity obligations for essential and important entities, applicable to AI operators in scope. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix EO 14110 (US Executive Order on Safe, Secure and Trustworthy AI) Federal direction on AI safety, civil rights, and innovation; tracked together with successor administration policy. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix NIST GenAI Profile NIST AI RMF profile addressing generative-AI-specific risks (hallucination, IP, bias, dangerous content, privacy). AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Colorado AI Act (SB24-205) Risk-management programme, consumer disclosures and impact assessments for high-risk AI systems. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix UK AI White Paper and sector-regulator follow-up Principles-based pro-innovation regulation administered through existing sector regulators. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Canada AIDA (Bill C-27) High-impact AI system risk-management, monitoring and accountability framework. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Singapore Model AI Governance Framework v2 Voluntary internal governance, human involvement, stakeholder communication and AI Verify toolkit. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix Brazil PL 2338/2023 (Marco Legal da IA) Risk-tiered AI regulation with fundamental-rights impact assessment and ANPD oversight. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix China Interim Measures for Generative AI Services Pre-deployment security assessments, content moderation and algorithm filings. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix ISO/IEC 23894:2023 AI risk management — guidance aligned with ISO 31000 applied to AI systems. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix ISO/IEC 5338:2023 AI system life cycle processes — process reference model for AI. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix ISO/IEC 38507:2022 Governance implications of the use of AI by organizations. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix VNG AI Governancekader (2025/2026) Dutch municipal AI governance framework: process, roles, knowledge bank, project register. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix BZK Algoritmekader 2.0 (2026) Dutch central-government operational reference for algorithm and AI governance. AI-BOK v1.1 AI-BOK v1.1, Regulatory Traceability Matrix AI Impact Assessment Assessment of impact and risks per AI system. Possibly personal data Confidential 10 years Mandatory for high-risk AI AI-BOK v1.2, Module 1, section 9.4 Level 1: Initial Ad hoc, no structured AI governance. AI-BOK v1.2, Module 3, section 3.4 Level 2: Repeatable Basic processes and roles defined. AI-BOK v1.2, Module 3, section 3.4 Level 3: Defined Organisation-wide standards and procedures. AI-BOK v1.2, Module 3, section 3.4 Level 4: Managed Quantitatively managed, continuous monitoring. AI-BOK v1.2, Module 3, section 3.4 Level 5: Optimised Continuous improvement, innovation-driven. AI-BOK v1.2, Module 3, section 3.4 Dimension: Governance Policy, decision structures, responsibilities. AI-BOK v1.2, Module 3, section 3.4 Dimension: Process Lifecycle processes, quality gates, feedback loops. AI-BOK v1.2, Module 3, section 3.4 Dimension: People Roles, competencies, training, culture. AI-BOK v1.2, Module 3, section 3.4 Dimension: Technology Tooling, platforms, infrastructure, automation. AI-BOK v1.2, Module 3, section 3.4 Dimension: Culture AI Literacy, innovation mindset, ethical awareness. AI-BOK v1.2, Module 3, section 3.4 AI Maturity Assessment Assessment of AI maturity across 5 dimensions and 5 levels. No personal data Confidential AI-BOK v1.2, Module 3, section 3.1 FRIA (Fundamental Rights Impact Assessment) Assessment of impact on fundamental rights. Likely personal data (fundamental rights assessment) Confidential 10 years Mandatory (Art. 27) AI-BOK v1.2, Module 3, section 3.1 DPIA for AI Data Protection Impact Assessment specifically for AI systems. Likely personal data (DPIA) Confidential 10 years (AVG) Mandatory (GDPR Art. 35) AI-BOK v1.2, Module 3, section 3.1 AI Quality Scorecard Quality score across 5 dimensions per AI system. No personal data Internal AI-BOK v1.2, Module 3, section 3.1 Risk Profile Beoordeling van risico's verbonden aan een AI-systeem. AI-BOK v1.2, Module 3, section 3.2 FM01 Prompt Injection (direct) Adversary inserts instructions in user input that override system instructions or extract privileged content. Mitigation: instruction-hierarchy enforcement; policy-consultation before privileged tool use. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM02 Indirect Prompt Injection (data-borne) Malicious instructions hidden in retrieved content treated as instructions. Mitigation: treat retrieved content as data; trust-graded knowledge sources. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM03 Tool Misuse Agent invokes legitimate tool with illegitimate parameters. Mitigation: per-tool capability gating; impact-threshold gates; policy-as-code evaluation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM04 Unauthorised Delegation Agent spawns sub-agent with inappropriate inherited or escalated privileges. Mitigation: explicit minted sub-mandates; chain-of-custody in authority register. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM05 Multi-agent Coordination Conflict Overlapping or contradictory mandates produce inconsistent state or oscillation. Mitigation: single-writer rule; mandate-intersection check at grant time. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM06 Memory Poisoning False or biased content inserted into persistent memory becomes ground truth. Mitigation: write-controls; trust-tiered memory; provenance and re-validation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM07 Context-Window Saturation Hostile content fills context, pushing system instructions out of effective attention. Mitigation: instruction re-injection at decision boundaries; structural separation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM08 Function-Call Abuse Malformed or malicious arguments to structured function calls. Mitigation: strict schemas; tenant-scope enforcement; function-side validation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM09 Goal Drift Effective objective diverges from stated goal over a long-running task. Mitigation: immutable objective binding; re-confirmation gates; alignment scoring. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM10 Sycophancy and Confidence Inflation Agent agrees with user assertions despite counter-evidence; humans accept without scrutiny. Mitigation: calibration; uncertainty channels; counter-evidence tests; KA13 literacy. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM11 Jailbreak Persistence Successful jailbreak establishes session/memory state that outlives the trigger. Mitigation: state reset on safety transitions; quarantined memory. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM12 NHI Credential Leak or Theft NHI credentials extracted via logs, prompts or compromised storage and reused. Mitigation: short-lived context-bound tokens; secret scanning; rotation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM13 Sandbox / Capability Escape Agent escapes execution sandbox to acquire capabilities outside its mandate. Mitigation: capability-based sandbox; deny-by-default network; ephemeral envs. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM14 Autonomy Creep Mandate gradually expanded across reviews without cumulative-risk re-assessment. Mitigation: full mandate review on change; cumulative-risk metric. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM15 Capability Inheritance via Tool Chaining Composite effects beyond any single tool's mandate. Mitigation: sequence-level policy evaluation; composite-mandate definitions. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM16 Hallucinated Authority Agent asserts approval/authority that does not exist; downstream actors trust. Mitigation: downstream verification of signed policy-consultation artefacts. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM17 Sensor Spoofing (operational) Adversary injects false sensor readings into operational AI. Mitigation: cross-sensor consistency; trust scoring at ingestion. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM18 Actuator Hijack (cyber-physical) Control plane breach actuates physical equipment outside cognition-plane scope. Mitigation: actuation-side authentication and rate-limiting; safe-state envelope. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM19 Closed-loop Drift (cyber-physical) Model and physical system co-drift; symptoms only visible after long horizons. Mitigation: bounded-deviation envelopes; ground-truth re-validation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM20 Sample-rate Mismatch (operational) Deployment sample rate differs from training; subtle in control loops. Mitigation: rate metadata; deployment-time validation. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM21 Connectivity-induced Policy Bypass (edge) Extended disconnection lets cached permissive policy outlive a revocation. Mitigation: mandate freshness budget; degraded mode. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue FM22 Hardware Degradation (operational) Sensor/accelerator drift; model accepts degraded inputs. Mitigation: hardware health telemetry; degradation-aware fallback. AI-BOK v1.1 Agentic AI failure mode AI-BOK v1.1, Failure-mode catalogue Data Quality Completeness, currency, representativeness, labelling, bias, semantic consistency. AI-BOK v1.2, Module 3, section 3.7 Model Quality Accuracy, robustness, reproducibility, generalisability, fairness. AI-BOK v1.2, Module 3, section 3.7 Interaction Quality Task success, user satisfaction, trust calibration, Hallucinationpercentage. AI-BOK v1.2, Module 3, section 3.7 Operational Quality Availability, reliability, scalability, drift detection speed, costs. AI-BOK v1.2, Module 3, section 3.7 Governance Quality Compliance ratio, auditability, transparency, ethical review coverage. AI-BOK v1.2, Module 3, section 3.7 Policy-as-code Governance Governance rules must be machine-readable and runtime-evaluable. Not Started Must Have Yes Immutable Audit Trail All AI decisions and interactions must be immutably logged. Not Started Must Have Yes Agent Authorisation and Mandate Agents must be authorised at runtime based on explicit mandates. Not Started Must Have Yes Input/Output Guardrails AI output must be filtered for safety, correctness and ethics. Not Started Must Have Yes Continuous Observability AI systems must be continuously monitored for performance, drift and costs. Not Started Must Have Yes Semantic Anchoring AI systems must interpret concepts via authoritative terminology frameworks. Not Started Must Have Yes End-to-end Traceability Every AI output must be traceable to sources, models and reasoning steps. Not Started Must Have Yes Central Model Management All production models must be registered with model cards and versions. Not Started Must Have Yes Structural Bias Detection Models must be periodically tested for undesirable biases. Not Started Must Have Yes Explainability of decisions Stakeholders have the right to an understandable explanation of AI decisions. Not Started Must Have Yes Privacy by design Personal data is never unnecessarily exposed to AI components. Not Started Must Have Yes Grounding in Authoritative Sources AI output must be based on verified, traceable knowledge sources. Not Started Must Have Yes Open standards and interoperability AI components communicate via standardised protocols (MCP, A2A). Not Started Must Have Yes Automated Delivery Pipeline Model-to-production must be maximally automated via CI/CD. Not Started Must Have Yes User-centred Prompt Design AI interactions are designed from the perspective of task, context and knowledge level. Not Started Must Have Yes Continuous Data Quality Monitoring Data quality is an ongoing process of measurement and improvement. Not Started Must Have Yes Board of Directors Strategic decision-making on AI deployment and risk acceptance. CIO / CDO Responsible for IT/data strategy and AI integration. Business Owner Owner of the business process supported by AI. End User Employee who works with AI systems daily. Citizen / Customer Person affected by AI decisions. Regulator External party overseeing AI Compliance (e.g. DPA, ACM). AI Team Multidisciplinary team of AI specialists. AI A technology that, for explicit or implicit purposes, infers how to generate outputs based on received input (such as data). Outputs include predictions, content, recommendations and/or decisions. The technology can learn, reason and perform tasks in a way that normally requires human intelligence. AI (Artificial Intelligence) refers to the broader concept of artificial intelligence - the science and technology that enables machines to mimic human cognitive tasks such as learning, reasoning and problem-solving. This includes techniques such as machine learning, neural networks and natural language processing. https://www.rijksoverheid.nl/documenten/publicaties/2025/04/22/overheidsbrede-handreiking-generatieve-ai kunstmatige intelligentie AI (artificial intelligence) is a technology where computers perform tasks normally done by humans, such as learning, thinking and problem-solving. AI can recognise faces, understand text or drive a car autonomously. This is done with smart programs that process large amounts of data and learn from them. AI-systeem A machine-based system designed to operate with varying levels of autonomy that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments. Examples of AI systems are machine learning, Natural Language Processing (NLP) and computer vision (classifiers). https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/definities/ kunstmatige intelligentie An AI system encompasses not only the model but also the entire infrastructure around it. This includes hardware, software, data processing, input and output interfaces, and all components needed to make the model work effectively. Examples are ChatGPT, Google Gemini and Co-Pilot. https://www.rijksoverheid.nl/documenten/publicaties/2025/04/22/overheidsbrede-handreiking-generatieve-ai Algorithm A set of rules and instructions that a computer automatically follows when making calculations to solve a problem or answer a question. https://www.rekenkamer.nl/binaries/rekenkamer/documenten/rapporten/2021/01/26/aandacht-voor-Algorithms/Aandacht+voor+Algorithms.pdf By algorithm we mean a set of rules and instructions that a computer executes, with goals such as solving problems, answering questions, executing tasks or making decisions. A recipe for solving a computer science problem from a given initial state to an intended end goal, consisting of a finite sequence of unambiguously defined instructions or rules. Model A program trained to recognise patterns in data and make predictions. concept For example, predicting weather, recognising cats. A model is a simplified representation of something from the real world. It helps to better understand, investigate or predict something. AI-model An AI model is the result of training an algorithm on data. An algorithm is a set of instructions, and the model is the specific result of following that set of instructions based on certain data. concept A model is trained on a task, for example classifying documents. GPT-4 is an example of an AI model. https://www.rijksoverheid.nl/documenten/publicaties/2025/04/22/overheidsbrede-handreiking-generatieve-ai Machine Learning The way computers learn new things without explicit programming, by learning from labelled data to make predictions. ML Self-learning algorithms are algorithms that train themselves. This process is called machine learning. This allows computers to learn from data without explicit programming. This is a very common form of AI. https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/soorten-Algorithms/#zelflerende-Algorithms Supervised Learning Machine learning process where a model learns from labelled examples (input-output pairs), comparable to a teacher-student scenario. gecontroleerd leren Your algorithm learns from data you label with information. For example, you offer photos with labels: this is a cat, this is not a cat. https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/soorten-Algorithms/#zelflerende-Algorithms Unsupervised Learning Machine learning proces waarbij AI zelf verborgen patronen in ongelabelde data zoekt zonder expliciete begeleiding. You let the algorithm discover patterns and structures itself in unstructured data without labels. https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/soorten-Algorithms/#zelflerende-Algorithms Reinforcement Learning The algorithm learns through punishment and reward. The goal is to score as high as possible in as little time as possible. For example, you give points when the algorithm sorts photos that look like cats. In reinforcement learning, the AI model learns autonomously. You can also choose to deploy the model frozen. https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/soorten-Algorithms/#zelflerende-Algorithms bekrachtiginsleren Validation Step in machine learning to verify model performance using a separate dataset (validation set) not used for training. Neural Networks An advanced form of AI that recognises complex patterns in data via layered neural networks, useful for image recognition, speech processing and language understanding. deeplearning CNN (Convolutional) Vooral gebruikt bij beeldherkenning. Convolutional Neural Networks RNN (Recurrent) Suitable for sequence data such as text and speech. Recurrent Neural Networks Transformers Backbone of modern NLP models such as GPT and BERT. Generative AI A form of artificial intelligence that can produce (generate) content (text, images and/or varied content such as music) based on the data it is trained on. GenAI An example is generating synthetic data and providing programming assistance. https://datanorth.ai/nl/blog/agentic-ai-begrijpen-definitie-en-toepassingen-in-de-praktijk Generative AI, often used by large language models such as GPT-4o and DALL-E 3, focuses on creating new content based on patterns learned from enormous datasets. Large Language Model A specialised type of generative AI model trained on large volumes of text to understand existing content and generate textual content. Large Language Model https://www.rijksoverheid.nl/documenten/publicaties/2025/04/22/overheidsbrede-handreiking-generatieve-ai Type of AI trained on large amounts of text data, capable of creating human-sounding text, answering questions, etc. GPT A type of neural network, pre-trained on large text data, that generates clear and relevant text based on prompts. Generative Pre-trained Transformer An example is ChatGPT. RAG A technique within artificial intelligence that combines the power of generative AI with information retrieval. This means an AI model not only generates text based on pre-trained data, but can also retrieve relevant information from external knowledge sources to provide more accurate and current answers. Retrieval Augmented Generation Prompt Engineering The process of carefully choosing and adapting the input (prompt) for a machine learning model to get the best possible output. prompten Examples of prompt engineering: asking clear questions, adding specific instructions for style, creating creative prompts for interesting ideas or designs. https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-prompt-engineering Prompt engineering is the art of smartly formulating instructions for an AI system. By using the right words and structure, an AI can give better and more accurate answers. Token The basic unit of text (a word or part of a word) processed by LLMs. Temperature The degree of randomness in the output of an LLM. Hallucination Output from generative AI that is factually incorrect or does not match reality, despite appearing semantically correct. Bias In the context of AI, bias refers to the assumptions an AI system makes to simplify the learning process and task execution. AI researchers try to minimise bias, as it can lead to poor results or unexpected outcomes. Bias in the context of AI refers to systematic errors or prejudices in an algorithm arising from biased training data. This can lead to unfair or inaccurate results. Bias can take various forms: Sample Bias, Measurement Bias, Algorithmic Bias. vooringenomenheid Natural Language Processing The way computers learn new things without explicit programming, by learning from labelled data to make predictions. Natural Language Processing NLP Named Entity Recognition Identifying names, locations and concepts in text. Named Entity Recognition Sentiment Analysis Interpreting whether a text is positive, negative or neutral. AI-agent An AI agent is a software entity that autonomously executes tasks and makes decisions based on observations, rules and machine learning algorithms. zelflerende bot AI agents can vary in complexity, from simple bots to advanced self-learning systems. https://news.microsoft.com/source/features/ai/ai-agents-what-they-are-and-how-theyll-change-the-way-we-work/ An AI agent is a computer program that can independently make decisions and take actions to achieve a goal. It uses AI to gather information, learn and adapt to changes. Agentic AI Agentic AI gives the AI system the ability to act autonomously, for example to create files. An example of Agentic AI is a digital assistant that not only answers questions but also independently plans a trip, books tickets and reserves a hotel based on preferences. https://datanorth.ai/nl/blog/agentic-ai-begrijpen-definitie-en-toepassingen-in-de-praktijk Agentic AI refers to artificial intelligence that can independently make decisions and take actions without direct human intervention. This type of AI goes beyond traditional AI systems. Agentic AI can plan, execute and adapt based on changing circumstances. Autonomous Agent These can independently execute complex tasks, such as self-driving cars that analyse traffic conditions and act without direct human control. Reactive Agent These react directly to input without long-term memory. For example, a chess computer that only analyses the current move. Multi-agent System These are multiple AI agents collaborating, for example robots in a factory jointly optimising a production process. EU AI Act The European AI Act is one of the world's first comprehensive laws specifically for artificial intelligence (AI). The AI Act establishes frameworks and requirements for both governments and businesses regarding the development and use of AI systems. https://www.digitaleoverheid.nl/overzicht-van-alle-onderwerpen/nieuwe-technologieen-data-en-ethiek/artificiele-intelligentie-ai/ai-verordening/ AI Act The purpose of the law is that AI systems used by organisations within the EU are safe and respect fundamental rights. AI systems are divided into different risk categories with corresponding rules. The AI Act is a European law that sets rules for the development and use of AI systems. The law also grants rights to citizens who come into contact with AI systems. AI Literacy Skills, knowledge and understanding enabling providers, deployers and affected persons, taking into account their respective rights and obligations under the AI Act, to make informed use of AI systems and to become more aware of the opportunities and risks of AI and the potential harm it can cause. https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/definities/ AI Detector An AI Detector is a tool designed to detect when a piece of text (or sometimes an image or video) has been created by AI tools (such as ChatGPT and DALL-E). These detectors are not 100% reliable, but they can provide an indication of the likelihood that a text was generated by AI. AI Detectors work with similar large language models (LLMs) as the tools they try to detect. They look for low levels of perplexity and burstiness. AI Detectors can be used by universities and other institutions to identify AI-generated content. https://www.scribbr.nl/ai-tools-gebruiken/ai-termen-begrippenlijst/ Deepfake AI-generated or manipulated image, audio or video material that resembles existing persons, objects, places, entities or events, and would be wrongly perceived by a person as authentic or truthful. https://minbzk.github.io/Algorithmkader/overhetAlgorithmkader/definities/ Algorithm Impact Assessment An Algorithm Impact Assessment is a tool for making trade-off decisions when deploying algorithms and artificial intelligence. https://www.rijksoverheid.nl/documenten/publicaties/2025/04/22/overheidsbrede-handreiking-generatieve-ai Examples of Algorithm Impact Assessments are the IAMA and the AIIA. AIIA Algorithm Register A register with descriptions of algorithmic applications that directly or indirectly have a societal and/or economic effect on citizens or society as a whole. http://begrippen.kadaster.nl/ar/id/concept/AlgorithmRegister Cognition Plane The architecture layer where AI systems reason, interpret and make decisions. Distinguished from the data and control planes by non-deterministic, context-dependent behaviour. The cognition plane is the core architecture innovation of the AI-BOK. It requires its own governance mechanisms such as mandates and reasoning boundaries. AI-BOK v1.2, Module 3, section 3.6 Grounding Anchoring AI output in verified, authoritative sources so that answers are factually correct and traceable. Grounding is one of the three foundations of the cognition plane. Without grounding, an LLM produces answers based on training data that may be outdated or incorrect. AI-BOK v1.2, Module 3, section 3.6.4 Semantic Precision Unambiguous, formally defined concepts that enable AI systems to interpret concepts correctly. First foundation of the cognition plane. Requires authoritative terminology frameworks (SKOS/NL-SBB). AI-BOK v1.2, Module 3, section 3.6.4 AI Governance The entirety of policy frameworks, decision structures, responsibilities and mandates with which an organisation governs AI systems. KA1 is the central knowledge area of the AI-BOK. AI Governance is embedded, not imposed - mechanisms are part of the architecture. AI-BOK v1.2, Module 1, section 1.1 Shadow AI Unapproved use of AI tools and services by employees outside the view of IT and governance. Shadow AI poses a growing risk. The AI-BOK recommends channeling (providing safe alternatives) over prohibiting. AI-BOK v1.2, Module 1, section 1.13 Drift Detection Monitoring shifts in data or model behaviour that degrade the performance of an AI system in production. Distinction between data drift (input distribution changes) and concept drift (input-output relationship changes). Trigger for the retraining loop. AI-BOK v1.2, Module 1, section 8.6 Guardrail Programmable boundaries on the input and output of AI models that prevent undesirable behaviour. Guardrails are architecturally anchored in the cognition plane. They form an additional security layer on top of model training. AI-BOK v1.2, Module 1, section 6.8 Agent Mandate Formal specification of the autonomy boundaries, escalation rules and context restrictions within which an AI agent may operate. Agent mandates are the governance instrument for agentic AI. They specify what an agent may do, may not do, and when a human must intervene. AI-BOK v1.2, Module 1, section 1.5 Quality Gate Formal decision point in the AI lifecycle where an AI initiative is assessed against quality criteria before proceeding to the next phase. Each quality gate has specific criteria, responsible roles and deliverables. Comparable to TOGAF ADM quality reviews. AI-BOK v1.2, Module 3, section 3.1 Model Card (begrip) Standardised documentation format describing what an AI model does, for whom, with what limitations and what performance. Based on Mitchell et al. (2019). Each model card contains: purpose, owner, training data, evaluation results, limitations, fairness assessment. AI-BOK v1.2, Module 1, section 6.4 Human-in-the-loop Design pattern where a human is actively involved in the decision-making process of an AI system and assesses every output before it is executed. HITL, menselijke controle Three variants: human-in-the-loop (every decision), human-on-the-loop (sampling), human-in-command (policy/strategy). AI-BOK v1.2, Module 1, section 11.6 Knowledge Graph A graph structure representing entities and their interrelationships as a knowledge base for AI systems. Knowledge Graph, KG Knowledge graphs form a core component of the data plane. They are used for GraphRAG, entity resolution and semantic search functions. AI-BOK v1.2, Module 1, section 5.7 Embedding A numerical vector representation of text, image or other data that captures semantic meaning in a continuous vector space. Embeddings are the foundation of RAG pipelines and semantic search. Embedding quality determines retrieval effectiveness. AI-BOK v1.2, Module 1, section 5.5 Fine-tuning Further training a foundation model on domain-specific data to improve performance for a specific task. Alternative to RAG when domain knowledge must be structurally incorporated into the model. More expensive and complex than prompting. AI-BOK v1.2, Module 1, section 6.6 Responsible AI The practice of designing, developing and deploying AI systems in a manner that is ethical, transparent, fair and responsible. Verantwoorde AI Responsible AI encompasses fairness, explainability, privacy, safety and human oversight. It is the overarching theme of KA11. AI-BOK v1.2, Module 1, section 11.1 Responsible AI Value Creation Measurable business value from AI systems within acceptable risk boundaries. Level 1: AI as Tool AI supports individual tasks, no autonomy. User initiates and controls fully. AI-BOK v1.2, Module 3, section 3.6.6 Level 2: AI in Workflows AI is integrated into business processes, automated steps with human supervision. AI-BOK v1.2, Module 3, section 3.6.6 Level 3: Digital Assistant AI conducts conversations, answers questions, assists with decisions. Limited autonomy. AI-BOK v1.2, Module 3, section 3.6.6 Level 4: Orchestration Multiple AI agents collaborate, coordinated by an orchestrator. High governance intensity. AI-BOK v1.2, Module 3, section 3.6.6 Level 5: Autonomous Agents AI agents act independently within mandates. Very high governance requirements. AI-BOK v1.2, Module 3, section 3.6.6 Level 6: Federated Collaboration Agents from different organisations collaborate. Maximum governance complexity. AI-BOK v1.2, Module 3, section 3.6.6 Phase 1: Quick Wins (0-3 months) Inventory existing AI initiatives, assemble AI Board, classify risk, create awareness. AI-BOK v1.2, Module 3, section 3.5 Phase 2: Foundation (3-12 months) Governance framework operational, lifecycle established, first models in registry, basic monitoring. AI-BOK v1.2, Module 3, section 3.5 Phase 3: Scale-up (12-24 months) Organisation-wide rollout, cognition plane operational, multi-agent pilots, full compliance. AI-BOK v1.2, Module 3, section 3.5 Phase 4: Optimisation (24+ months) Continuous improvement, federated collaboration, AI-driven innovation, maturity level 4-5. AI-BOK v1.2, Module 3, section 3.5 Gap: Geen → Quick Wins Difference between current situation and first quick wins. Gap: Quick Wins → Fundament Difference between ad-hoc and structured. Gap: Fundament → Opschaling Difference between basic and organisation-wide rollout. Semantic Precision + Reliable Knowledge Sources + Explicit Governance zijn alle drie nodig. Cognition Plane Reasoning, interpretation, decision-making by AI. AI-BOK v1.2, Module 3, section 3.6.1 Control Plane Identity, authorisation, logging, audit. AI-BOK v1.2, Module 3, section 3.6.1 Data Plane Data, knowledge, vector stores, knowledge graphs. AI-BOK v1.2, Module 3, section 3.6.1 1st Line: Execution AI Engineers, AI Operations Engineers — dagelijkse operatie en eerste kwaliteitscontrole. AI-BOK v1.2, Module 3, section 3.3.4 2nd Line: Oversight AI Risk Manager, AI Compliance Officer — onafhankelijke controle en risicobewaking. AI-BOK v1.2, Module 3, section 3.3.4 3rd Line: Assurance AI Auditor - independent audit of governance and systems. AI-BOK v1.2, Module 3, section 3.3.4 LLM Providers Agent & RAG Guardrails & Safety Evaluation & Fairness Prompt Management Identity & Access Policy Engines MLOps & Registry API Gateways Monitoring CI/CD & Infra Model Serving Vector Databases Knowledge Graphs Data Platforms Data Quality Feature Stores Annotation Document Storage Data Lineage Cloud & Research Architecture Management Plane Management plane: configuration, deployment, monitoring, model registry, lifecycle and resourcing of AI systems. Added in line with the EA4AI paper (CBI-EDOC 2026, Springer LNBIP), which positions the cognition plane alongside the data, control and management planes. establishes governance frameworks for defines architecture principles determines gate criteria sets risk tolerance mandates compliance anchors ethics sets knowledge source requirements provides architecture requirements determines initiatives formulates data requirements prioritises model development determines operational capacity defines lifecycle blueprint specifies data access defines infrastructure requirements specifies interaction patterns provides target architecture specifies RAG architecture provides lifecycle framework provides systems for deployment provides training/evaluation data provides concept frameworks provides bias analyses provides data lineage provides bias detection provides semantic building blocks provides vamemberated models escalates model risks provides model cards provides fairness assessments provides interaction patterns provides monitoring data escalates incidents provides operational logs provides risk input shares risk-ethics interface shares FRIA assessment provides context specifications provides source references provides hallucination prevention provides traceability chairperson member member member member Retraining loop Evaluation loop Evolution loop involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in involved in supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by supported by primary primary primary primary primary primary primary primary primary primary applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires requires leads has interest in has interest in has interest in has interest in has interest in has interest in has interest in applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to applies to provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides provides trained on receives consults results in has falls under generates contains governs governs governs governance governance governance governance governance governance governance governance performance degradation retrain evaluation failed retrain fundamental change restart cycle has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept has as concept reports into and informs Introduced in AI-BOK v1.1 (KA13). anchors Article 4 evidence in Introduced in AI-BOK v1.1 (KA13). co-owns ethical-use content with Introduced in AI-BOK v1.1 (KA13). derives supervision literacy from Introduced in AI-BOK v1.1 (KA13). paired with (human grounding ↔ machine grounding) Introduced in AI-BOK v1.1 (KA13). supplies a documented control to Introduced in AI-BOK v1.1 (KA13). accountable for AI Literacy Officer is the accountable role for KA13. registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in registered in cognition-plane binding registered in registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding registered in cognition-plane binding requires compliance through AI Act Article 4 is operationalised by KA13. sector implementation layer of sector implementation layer of standard anchor for standard anchor for policy consultation / audit trail access control / lineage configure / monitor configure / monitor configure / monitor ai-act-risk-class automation-level availability-sla business-value category compliance-relevance confidentiality cost-indication data-sensitivity dct:source deployment-model enterprise-readiness eu-ai-act-relevance eu-ai-act-relevant fte-indication gdpr-classification governance-intensity human-oversight implementation-status integration-protocol introduced-in ka-ring license-model lifecycle-phase maturity-level origin plane priority raci-default-for rdfs:comment retention-period role-number seniority-level skos:altLabel skos:closeMatch skos:notation skos:scopeNote url vendor-dependency vendor-type 01.01 AI Capability Map (13 Knowledge Areas) Source: AI-BOK v1.2, Module 1, KA1 to KA13 AI-BOK v1.2, Module 1, KA1 to KA13 01.02 Layered View (all layers) Source: AI-BOK v1.2, Module 3, section 3.6 AI-BOK v1.2, Module 3, section 3.6 01.00 AI-BOK Overview - Why, What, How & Who Source: AI-BOK v1.2, Module 1, KA1 to KA13 AI-BOK v1.2, Module 1, KA1 to KA13 02.01 AI Role Model (20 Roles) Source: AI-BOK v1.2, Module 2, section 2.2 AI-BOK v1.2, Module 2, section 2.2 02.02 AI Lifecycle Management (TOGAF ADM style) Source: AI-BOK v1.2, Module 1, section 4 AI-BOK v1.2, Module 1, section 4 02.03 Process Model with Quality Gates Source: AI-BOK v1.2, Module 1, section 4 AI-BOK v1.2, Module 1, section 4 02.04 RACI Matrix (Roles x Knowledge Areas) 02.05 Business Functions & Business Objects (52 objects) Source: AI-BOK v1.2, Module 1, KA1 to KA13 AI-BOK v1.2, Module 1, KA1 to KA13 02.06 Function Model with ABBs Source: AI-BOK v1.2, Module 3, section 3.6.5a AI-BOK v1.2, Module 3, section 3.6.5a 02.10 AI Value Stream Source: AI-BOK v1.2, Module 3, section 3.5 AI-BOK v1.2, Module 3, section 3.5 02.07 Three Lines of Defence Source: AI-BOK v1.2, Module 3, section 3.3.4 AI-BOK v1.2, Module 3, section 3.3.4 02.08 Sub-roles and Specialisations Source: AI-BOK v1.2, Module 2, section 2.2 AI-BOK v1.2, Module 2, section 2.2 02.09 Business Objects per Knowledge Area Source: AI-BOK v1.2, Module 1, KA1 to KA13 AI-BOK v1.2, Module 1, KA1 to KA13 03.00 Planes Overview (incl. Management Plane) The four architectural planes: the cognition plane (mandates, reasoning boundaries, grounding) alongside the control plane (policy, audit), data plane (sources, RAG) and management plane (configuration, monitoring, lifecycle). Grounding is the assurance relation between reasoning and its sources; the sources themselves reside in the data plane. Detailed ABBs per plane: views 03.01 to 03.04. 03.01 Planes Architecture (ABBs + Services) Source: AI-BOK v1.2, Module 3, section 3.6.1 See 03.00 for the planes overview including the management plane. AI-BOK v1.2, Module 3, section 3.6.1 03.02 SBBs Cognition Plane Source: AI-BOK v1.2, Module 3, section 3.6.1 AI-BOK v1.2, Module 3, section 3.6.1 03.03 SBBs Control & Management Plane Source: AI-BOK v1.2, Module 3, section 3.6.1 AI-BOK v1.2, Module 3, section 3.6.1 03.04 SBBs Data Plane Source: AI-BOK v1.2, Module 3, section 3.6.1 AI-BOK v1.2, Module 3, section 3.6.1 03.05 AI Meta-model (Data Model) Source: AI-BOK v1.2, Module 3, section 3.2 AI-BOK v1.2, Module 3, section 3.2 03.06 Application Cooperation (Data Flows) Source: AI-BOK v1.2, Module 3, section 3.6.5 AI-BOK v1.2, Module 3, section 3.6.5 03.07 RAG Architecture Variants Source: AI-BOK v1.2, Module 1, section 12.6 AI-BOK v1.2, Module 1, section 12.6 03.08 Privacy Architecture (Key Vault Pattern) Source: AI-BOK v1.2, Module 1, section 5.10 AI-BOK v1.2, Module 1, section 5.10 03.09 Vendor Landscape Source: AI-BOK v1.2, Module 3, section 3.6.1 AI-BOK v1.2, Module 3, section 3.6.1 04.01 AI Governance Structure Source: AI-BOK v1.2, Module 1, section 1.2 AI-BOK v1.2, Module 1, section 1.2 04.00 Stakeholder Analysis Source: AI-BOK v1.2, Module 2, section 2.3 AI-BOK v1.2, Module 2, section 2.3 04.02 Principles Detail (KA1 Governance + KA3 Architecture) Source: AI-BOK v1.2, Module 1, section 1.2, 3.2 AI-BOK v1.2, Module 1, section 1.2, 3.2 04.03 All Principles Overview (72+) Source: AI-BOK v1.2, Module 1, KA1 to KA13 AI-BOK v1.2, Module 1, KA1 to KA13 04.08 Requirements Realisation (Principle → Requirement → ABB) Source: AI-BOK v1.2, Module 3, section 3.6.5a AI-BOK v1.2, Module 3, section 3.6.5a 04.04 Standards & Constraints (with ABB mappings) Source: AI-BOK v1.2, Module 3, section 3.3 AI-BOK v1.2, Module 3, section 3.3 04.05 AI Concepts in Context Source: AI Thesaurus (AI-BOK appendix) AI Thesaurus (AI-BOK bijlage) 04.07 Additional Standards & Environmental Factors Source: AI-BOK v1.2, Module 3, section 3.3 AI-BOK v1.2, Module 3, section 3.3 04.06 Governance Integration (TOGAF/DAMA/COBIT/GDPR) Source: AI-BOK v1.2, Module 3, section 3.3 AI-BOK v1.2, Module 3, section 3.3 04.09 All AI Concepts (Thesaurus) Source: AI Thesaurus (AI-BOK appendix) AI Thesaurus (AI-BOK bijlage) 05.01 Agent Maturity Spectrum (6 levels) Source: AI-BOK v1.2, Module 3, section 3.6.6 AI-BOK v1.2, Module 3, section 3.6.6 05.02 Maturity Model (5x5 with Roadmap & Agent Maturity) Source: AI-BOK v1.2, Module 3, section 3.4 AI-BOK v1.2, Module 3, section 3.4 05.03 AI Quality Framework (5 dimensions) Source: AI-BOK v1.2, Module 3, section 3.7 AI-BOK v1.2, Module 3, section 3.7 05.04 Implementation Roadmap (4 phases with Gaps) Source: AI-BOK v1.2, Module 3, section 3.5 AI-BOK v1.2, Module 3, section 3.5 06.01 KA1: AI Governance Source: AI-BOK v1.2, Module 1, section 1 AI-BOK v1.2, Module 1, section 1 06.02 KA2: AI Strategy & Portfolio Management Source: AI-BOK v1.2, Module 1, section 2 AI-BOK v1.2, Module 1, section 2 06.03 KA3: AI Architecture Source: AI-BOK v1.2, Module 1, section 3 AI-BOK v1.2, Module 1, section 3 06.04 KA4: AI Lifecycle Management Source: AI-BOK v1.2, Module 1, section 4 AI-BOK v1.2, Module 1, section 4 06.05 KA5: Data & Semantics for AI Source: AI-BOK v1.2, Module 1, section 5 AI-BOK v1.2, Module 1, section 5 06.06 KA6: Model Management Source: AI-BOK v1.2, Module 1, section 6 AI-BOK v1.2, Module 1, section 6 06.07 KA7: AI Interaction & User Experience Source: AI-BOK v1.2, Module 1, section 7 AI-BOK v1.2, Module 1, section 7 06.08 KA8: AI Operations Source: AI-BOK v1.2, Module 1, section 8 AI-BOK v1.2, Module 1, section 8 06.09 KA9: AI Risk Management & Safety Source: AI-BOK v1.2, Module 1, section 9 AI-BOK v1.2, Module 1, section 9 06.10 KA10: AI Compliance & Audit Source: AI-BOK v1.2, Module 1, section 10 AI-BOK v1.2, Module 1, section 10 06.11 KA11: AI Ethics & Responsible AI Source: AI-BOK v1.2, Module 1, section 11 AI-BOK v1.2, Module 1, section 11 06.12 KA12: AI Knowledge & Context Management Source: AI-BOK v1.2, Module 1, section 12 AI-BOK v1.2, Module 1, section 12 06.13 KA13: AI Literacy & Capability Development KA13 capability with AI Literacy Officer (Role 20), the related KAs (KA1, KA10, KA11, KA3, KA12, KA9) and the EU AI Act Article 4 anchor. Introduced in AI-BOK v1.1. AI-BOK v1.1 07.01 Agentic AI Failure-Mode Catalogue Sixteen agentic AI failure modes (FM01-FM16) plus six operational/cyber-physical variants (FM17-FM22), with their primary KA anchors (KA9 Risk, KA3 Architecture). Cognition-plane bindings per pattern documented in the v1.1 failure-mode catalogue. Introduced in AI-BOK v1.1. AI-BOK v1.1 08.01 Regulatory Landscape (multi-jurisdiction) Standards and constraints since v1.1: EU adjacent instruments (DSA, DMA, Data Act, NIS2), non-EU regimes (US, UK, Canada, Singapore, Brazil, China), ISO companions (23894, 5338, 38507), and Dutch sector instruments (VNG, BZK). Mapped to KA1/KA9/KA10/KA13 in the v1.1 regulatory traceability matrix. Introduced in AI-BOK v1.1. AI-BOK v1.1 09.01 GEMMA-style Local Register Pattern Pattern for embedding AI-BOK reference objects in a municipality/government core information-architecture register, alongside GEMMA application reference-components. Each AI system in the local algoritmeregister references the AI-BOK metamodel objects (AI System, AI Model, Dataset, Prompt, Interaction, Decision, Knowledge Source, Agent, Governance Framework, Risk Profile). Introduced in AI-BOK v1.1. AI-BOK v1.1