COMPEL Glossary — AI Transformation Terminology

The 18-Domain Maturity Model is the COMPEL assessment framework that evaluates an organization's AI capabilities, practices, and governance across 18 specific domains organized under the Four Pillars of People, Process, Technology, and Governance.

A/B testing is a controlled experiment that compares two versions of an AI model, interface, or process by exposing each to a different group of users and measuring which performs better against predefined metrics.

Absorption capacity is an organization's ability to recognize valuable new knowledge and technology, assimilate it into existing operations and mental models, and apply it productively to create business value.

Accountability in AI governance means that when an AI system causes harm, there are clear lines of human responsibility.

An accountability framework is a structured system that defines who is responsible for AI decisions, how those decisions are documented, what oversight mechanisms exist, and what consequences apply when things go wrong.

Accuracy is a model performance metric measuring the proportion of all predictions (both positive and negative) that are correct.

Action research is a cyclical research methodology where practitioners simultaneously study and improve their own professional practice by iterating through cycles of planning, acting, observing, and reflecting.

The action space is the complete set of all actions an AI agent can potentially take, including tool invocations (API calls, database queries, file operations), communication actions (messages to humans or other agents), reasoning actions (internal processing steps), and environmental interactions (network requests, system operations).

An adaptive learning system modifies its own behavior based on new data without explicit reprogramming, raising governance challenges as decision processes change over time.

Adaptive management is a structured, iterative approach to decision-making that explicitly acknowledges uncertainty and adjusts plans based on new information, monitoring results, and changing conditions rather than rigidly following an original blueprint.

Adoption rate measures the percentage of intended users who are actively and effectively using an AI-enabled tool or process.

A structured assessment of actual user adoption rates, workflow integration effectiveness, and change management outcomes — measuring whether end users are actually using the AI system as designed and whether the workflow redesign is delivering its intended effects.

The adoption trap is a COMPEL-identified anti-pattern describing the illusion of progress created by accumulating AI tools and deployments without building underlying organizational capability.

An adversarial attack is a deliberate attempt to fool or manipulate an AI system by providing specially crafted inputs designed to cause incorrect outputs.

An advisory board is a group of external experts and thought leaders who provide non-binding strategic guidance to an organization's AI transformation program, offering perspectives that internal teams may lack.

An advisory engagement is a COMPEL consulting arrangement where the practitioner provides ongoing strategic counsel to client leadership, guiding their internally-led transformation efforts without taking direct delivery responsibility.

A formal classification of every AI agent in scope according to its autonomy level — from level 0 (no autonomy, human executes) through level 4 (full autonomy, agent executes without human involvement) — with corresponding governance requirements, approval boundaries, and monitoring obligations assigned at each level.

The specialized governance framework for autonomous and semi-autonomous AI agents that extends traditional AI governance with agent-specific controls: autonomy level classification, tool access controls, data access boundaries, approval boundaries, human-in-the-loop thresholds, auditability requirements, fallback and kill switch mechanisms, escalation rules, simulation testing requirements, and agent risk tiering..

Agent lifecycle management is the end-to-end governance process covering the creation, registration, testing, deployment, monitoring, updating, and eventual retirement of AI agents within an enterprise.

Agent orchestration is the coordination of multiple AI agents working together on complex tasks, managing how work is routed between agents, how handoffs occur, how conflicts are resolved, and how the collective output remains coherent.

An agent registry is a centralized catalog that tracks all deployed AI agents across an organization, documenting their capabilities, permissions, owners, operational status, tool access, and governance compliance.

Agentic AI refers to artificial intelligence systems capable of taking autonomous actions in the world, making decisions, using external tools, and pursuing multi-step goals with minimal or no human intervention at each step.

Agentic AI transformation strategy is a comprehensive approach to deploying autonomous AI agents within enterprise transformation contexts, encompassing agent architecture design, autonomy level governance (from advisory through fully autonomous), human-in-the-loop oversight patterns, multi-agent orchestration frameworks, kill switch and containment protocols, observability and monitoring infrastructure, and compliance evidence collection for regulatory requirements.

An agentic failure taxonomy is a structured classification system that categorizes the types of failures that can occur in agentic AI systems, providing a shared vocabulary for identifying, discussing, and governing AI agent risks.

Agile is a set of principles for software development emphasizing iterative delivery, team collaboration, responsiveness to change, and working outputs over comprehensive documentation.

AI adoption is the act of introducing AI technologies into an organization's operations -- purchasing AI software, deploying pre-built models, or building custom solutions for specific use cases.

A concise, board-level declaration that defines the organization's strategic intent for AI transformation — what it aims to achieve, at what pace, and with what risk appetite.

An AI Capability Center is an organizational unit that concentrates AI expertise, tools, and shared resources to serve the broader enterprise, representing an evolution beyond the traditional Center of Excellence (CoE) model.

AI controls are the specific technical, procedural, and organizational mechanisms that enforce AI governance policies in practice.

An AI Demand Review Board is a governance body responsible for evaluating, prioritizing, and approving incoming AI project requests from across the organization, ensuring each initiative aligns with the enterprise AI strategy and that limited technical resources are allocated to the highest-value opportunities.

AI due diligence is the comprehensive investigation and assessment of AI capabilities, risks, liabilities, and technical assets conducted during mergers, acquisitions, partnerships, or major vendor selections.

An AI Ethics Board is a cross-functional body with genuine authority to review, approve, pause, or halt AI initiatives based on ethical criteria.

AI FinOps (Financial Operations for AI) is the practice of managing and optimizing the financial costs of AI infrastructure, including cloud compute spending, model training expenses, inference costs, data storage, and third-party API usage.

AI governance is the system of policies, roles, processes, oversight bodies, and controls that an organization uses to manage AI systems responsibly across their full lifecycle.

AI Incident Classification is a systematic framework for categorizing AI failures, malfunctions, and harmful outputs by their severity, impact scope, root cause type, and urgency of required response.

AI literacy is the degree to which individuals across an organization understand AI concepts, capabilities, and limitations well enough to make informed decisions within their domain and participate meaningfully in AI-enabled work.

AI maturity is the measured level of organizational capability in adopting, governing, and scaling artificial intelligence across all relevant dimensions — people, process, technology, and governance.

An AI operating model defines how an organization structures its people, processes, data, and technology to deploy and govern AI at scale.

A comprehensive design document that defines how the organization will govern, fund, staff, and operate AI capabilities at scale — covering the Center of Excellence structure, decision rights, team topologies, tooling standards, and operating procedures.

AI operating model readiness measures an organization's preparedness to establish and sustain the governance structures, decision rights, roles, committees, and processes required to operate AI systems at scale.

A structured, repeatable management system that enables an organization to plan, govern, deliver, measure, and continuously improve AI capabilities across people, process, technology, and governance dimensions.

AI Platform Strategy is the enterprise-level approach to selecting, building, and integrating the technology foundation that supports all AI development, deployment, and operations across the organization.

An AI product manager is a professional responsible for defining AI use cases, managing stakeholder engagement, translating business requirements into technical specifications, and ensuring that AI solutions deliver measurable business value.

An AI readiness assessment is a structured diagnostic that evaluates an organization's preparedness to adopt, govern, and scale AI across key dimensions: leadership alignment, data quality, technical infrastructure, workforce skills, governance frameworks, and regulatory posture.

AI Risk Champions are designated individuals embedded within business units who serve as local advocates for AI risk awareness and act as liaisons between frontline operations and the central AI risk management function.

An AI Risk Governance Board is a senior leadership body responsible for overseeing AI-related risks across the entire enterprise, establishing the organization's AI risk appetite, making decisions about acceptable risk levels for AI deployments, and ensuring that risk management practices are adequate and consistently applied.

An AI Risk Register is a documented, maintained inventory of all identified AI-related risks within an organization, capturing each risk's description, likelihood, potential impact, current mitigation measures, assigned owner, and review status.

AI Safety is the field of research and practice dedicated to ensuring that AI systems operate without causing unintended harm to individuals, organizations, or society.

AI Security Architecture is the comprehensive design of security controls and defense mechanisms specifically tailored to the unique threat landscape of AI systems, covering model protection against extraction and poisoning, training data security, adversarial input defense, prompt injection prevention, API access control, supply chain security for AI components, and audit trail integrity.

AI Service Level Management is the practice of defining, measuring, monitoring, and maintaining agreed-upon performance standards for AI services, extending traditional ITIL service management concepts to cover AI-specific metrics such as model accuracy, prediction latency, fairness consistency, drift thresholds, and retraining frequency.

The AI Steering Committee is the senior governance body that provides strategic direction, resolves cross-functional conflicts, approves budgets, and maintains executive accountability for AI transformation outcomes.

A formal register that classifies every AI system in scope according to risk tier, autonomy level, data sensitivity, regulatory applicability, and criticality — producing a system-level risk profile that determines which governance controls, review processes, and compliance requirements apply.

An AI System Impact Assessment is a structured, documented evaluation of how a proposed or existing AI system affects individuals, groups, organizations, and society across dimensions including fundamental rights, safety, privacy, fairness, environmental impact, and labor market effects.

An AI System Registry is an organizational catalog that documents all AI systems in use, under development, or being evaluated, recording each system's purpose, data inputs, risk classification, ownership, compliance status, deployment environment, and review history.

AI transformation is the enterprise-wide process of systematically adopting, governing, and scaling artificial intelligence to change how an organization operates, competes, and creates value.

A specific, bounded application of AI technology to a defined business process or decision point, with a documented value thesis, identified stakeholders, risk classification, and success criteria.

AIOps (Artificial Intelligence for IT Operations) is the application of AI and machine learning to IT operations tasks such as monitoring, anomaly detection, alerting, root cause analysis, and automated incident resolution.

AITF is the Level 1 COMPEL certification demonstrating foundational mastery of the COMPEL methodology, including the six lifecycle stages (Calibrate, Organize, Model, Produce, Evaluate, Learn), the Four Pillars (People, Process, Technology, Governance), the 18-domain maturity model, and the core principles of AI transformation.

AITGP is the Level 3 COMPEL certification for professionals who have mastered the skills needed to architect enterprise-level AI transformation strategies, design operating models, build governance frameworks, and mentor specialist-level practitioners.

AITP is the Level 2 COMPEL certification for practitioners who can independently design, lead, and deliver COMPEL transformation engagements with real client organizations, including advanced maturity assessment, transformation roadmap architecture, multi-workstream execution management, and measurement and evaluation.

AITP Lead is the Level 4 apex certification in the COMPEL framework for professionals who govern portfolios of AI transformation programs across multiple organizations, harmonize governance across organizational boundaries, design enterprise AI operating models, and actively contribute to industry standards development.

Agentic Transformation Specialist.

AI Solution Architecture for Transformation Leaders.

Value-Driven Transformation.

Workforce Change for Transformation.

An algorithm is a set of step-by-step instructions or mathematical rules that a computer follows to solve a problem or complete a task.

Algorithmic accountability is the principle that organizations deploying algorithms must be answerable for the outcomes those algorithms produce, including unintended consequences, discriminatory effects, and errors that affect individuals.

An algorithmic audit is an independent, systematic examination of an AI system's decision-making processes, data inputs, outputs, and real-world impacts to assess whether the system operates in compliance with legal requirements, ethical standards, and organizational policies.

Algorithmic bias is systematic and unfair discrimination in AI system outputs, often arising from biased training data, flawed model design, unrepresentative data samples, or proxy variables that encode protected characteristics.

An Algorithmic Impact Assessment (AIA) is a formal, structured evaluation conducted before deploying an AI system to identify and quantify potential negative impacts on individuals and communities, particularly regarding fairness, privacy, civil rights, employment, and access to services.

Andragogy is the theory and practice of adult education, distinct from pedagogy (child education), recognizing that adults learn differently and have specific needs including understanding why they are learning something, drawing on their existing experience, exercising self-direction, focusing on immediately applicable knowledge, and being motivated by internal rather than external factors.

Anomaly detection is a technique that identifies data points, events, or patterns that deviate significantly from expected behavior.

Anonymization is the process of irreversibly removing or altering personally identifiable information from datasets so that individuals cannot be re-identified, even by combining the anonymized data with other available information.

In AI transformation, an anti-pattern is a commonly occurring organizational behavior that appears rational in the moment but systematically undermines transformation outcomes.

An API is a set of rules and protocols that allows different software systems to communicate with each other in a standardized way.

An Architecture Review Board (ARB) is a governance body that evaluates proposed technology designs, platform changes, and system integrations against enterprise architecture standards, ensuring consistency, scalability, security, and strategic alignment before implementation begins.

In the COMPEL framework, an artifact is a formal document, record, or deliverable produced during the lifecycle that provides evidence of governance activities, decisions, and outcomes.

Artificial General Intelligence is a theoretical form of AI with human-level cognitive ability across all intellectual domains -- the ability to reason, learn, and adapt to any task a human can perform.

Artificial Intelligence is a broad field of computer science focused on building systems that can perform tasks typically requiring human intelligence, such as understanding language, recognizing patterns, making decisions, and generating content.

Artificial Narrow Intelligence describes AI that performs a specific task within a defined domain, such as detecting fraud, translating languages, or playing chess.

An assessment-only engagement is a COMPEL consulting arrangement focused exclusively on diagnosing an organization's AI maturity using the 18-domain maturity model, typically lasting four to eight weeks with a smaller team and lower commercial risk than full transformation engagements.

Assurance is the process of providing justified confidence to stakeholders that AI systems, processes, and governance mechanisms are operating effectively, safely, and in compliance with stated requirements and standards.

The attack surface of an AI system encompasses all the points where an unauthorized actor could attempt to access, manipulate, or extract data from the system, including model API endpoints, training data pipelines, model weight storage, user interfaces, supply chain components, and the inference process itself.

Attestation is a formal declaration by an authorized person or body that an AI system, process, or governance practice meets specified requirements or standards at a particular point in time.

Audit preparedness is the continuous operational discipline of ensuring that governance activities produce the documentation, evidence trails, and records that auditors and regulators can verify on demand.

An audit trail is a chronological record of all activities, decisions, and changes related to an AI system, maintained to support accountability, compliance verification, and regulatory examination.

Auto-scaling is the automatic adjustment of computing resources, such as servers, containers, or GPU instances, based on real-time demand patterns.

Autonomy calibration determines appropriate AI autonomy levels for specific tasks, balancing efficiency against risks and regulatory requirements.

The autonomy spectrum is a classification framework that describes how independently an AI agent can operate, ranging from Level 0 (no autonomy -- fully deterministic instructions) through Level 5 (full autonomy -- self-directed goal setting).

The visual and structural tier assigned to a credential in the lattice, indicating its relative level.

A Balanced Scorecard is a strategic performance measurement framework that tracks metrics across four complementary perspectives: financial results, customer satisfaction, internal process efficiency, and organizational learning and growth.

A baseline is a documented measurement of current performance, capability, or conditions taken before an AI initiative begins, providing the reference point against which progress, improvement, and ROI are measured.

Batch inference is the practice of running an AI model's predictions on a large collection of data items simultaneously, rather than processing them one at a time in real time.

Batch processing involves running AI model predictions on large volumes of data at scheduled intervals rather than in real time.

A benchmark is a standardized test, dataset, or reference point used to evaluate and compare AI model performance against a common standard.

A report that updates the organization's AI performance and maturity benchmarks based on actual results from the current evaluation cycle — revising baselines, adjusting target ranges, and incorporating new external benchmarks to ensure that future KPI targets remain meaningful and appropriately ambitious.

A benefits register catalogs expected and realized benefits specifying description, magnitude, measurement method, owner, and evidence.

Benefits tracking is the systematic, ongoing process of measuring and documenting the actual value delivered by an AI transformation program against the projected benefits that justified the original investment.

Bias auditing is the systematic review of AI training data and model outputs to identify and measure unfair biases.

Bias detection is the process of systematically identifying unfair patterns in AI systems, examining training data for historical prejudices, model outputs for discriminatory patterns, and real-world impacts for disproportionate effects on particular demographic groups.

Binding Corporate Rules (BCRs) are internal data protection policies adopted by multinational organizations and approved by data protection authorities that allow the lawful transfer of personal data between entities within the corporate group across different countries.

A blameless post-mortem is an incident review methodology that deliberately focuses on understanding systemic causes, process failures, and improvement opportunities rather than assigning personal blame to individuals involved in an AI system failure or incident.

Bloom's Taxonomy is a hierarchical framework for classifying educational learning objectives into six levels of increasing cognitive complexity: Remember, Understand, Apply, Analyze, Evaluate, and Create.

Board-level governance refers to the oversight, strategic direction, and fiduciary responsibility that an organization's board of directors exercises over AI transformation, including setting risk appetite for AI initiatives, approving AI strategy, allocating transformation investment, and holding the executive team accountable for responsible AI practices.

A Body of Knowledge (BoK) is the complete, structured set of concepts, terms, theories, practices, tools, and standards that define a professional discipline and are required for competent practice within it.

A bridge credential is a credential type specifically designed to connect external technical training (such as partner bootcamps in data science, LLM engineering, or agentic AI) with COMPEL transformation methodology.

The Brussels Effect describes the tendency of European Union regulation to become the de facto global standard because multinational organizations find it more efficient to adopt a single, stringent standard globally than to maintain different compliance practices for different jurisdictions.

Buffer management is the deliberate practice of building time and resource margins into project schedules and dependency chains to absorb inevitable delays, unexpected complications, and minor failures without triggering cascading schedule disruptions across connected workstreams.

A business case is a structured document or analysis that provides the financial and strategic justification for an AI investment by quantifying expected costs, benefits, risks, timelines, and alternative options.

Business continuity planning ensures that an organization's critical functions can continue during and after disruptions, including AI system failures, data breaches, infrastructure outages, and vendor collapses.

C-suite advisory is the practice of providing strategic counsel to an organization's most senior executives, including the CEO, CTO, CFO, CISO, CDO, and other C-level leaders, on AI transformation strategy, organizational implications, risk landscape, and investment priorities.

Calibrate is the first of the six COMPEL stages, focused on producing an honest, evidence-based assessment of where the organization stands today in its AI transformation journey.

The Calibrate stage is the first stage of the COMPEL lifecycle where an organization's current AI maturity is systematically assessed across the 18-domain maturity model, establishing a data-driven baseline that informs all subsequent transformation planning.

Canary deployment is a risk-mitigation release strategy where a new version of an AI model or system is first deployed to a small, carefully selected subset of production traffic, and its performance is monitored closely before gradually expanding the rollout to the full user base.

A grounded inventory of the AI, data, platform, and talent capabilities currently in place across the organization, captured at a point in time.

Capability compounding is the principle that AI capabilities build upon each other, with each making subsequent ones easier and more valuable.

CMMI is a process improvement framework originally developed at Carnegie Mellon University that defines maturity levels for organizational processes.

Capital allocation is the strategic process of distributing financial resources across a portfolio of AI transformation initiatives based on strategic priorities, expected returns, risk profiles, and organizational readiness.

The capstone portfolio is the comprehensive collection of artifacts, analyses, strategy documents, governance frameworks, and reflective narratives that Level 4 AITP Lead candidates assemble to demonstrate mastery across portfolio leadership, cross-organizational governance, operating model design, framework interoperability, and industry standards contribution.

A capstone project is a comprehensive, integrative assessment in the COMPEL certification program where candidates must demonstrate mastery by applying the full methodology to a real or simulated enterprise scenario.

A cascading failure is a chain reaction where one component's malfunction triggers failures in dependent components, which in turn cause further failures, potentially resulting in widespread or total system collapse.

Catastrophic forgetting occurs when AI models trained on new data lose previously acquired knowledge.

The California Consumer Privacy Act (CCPA) is a US state data privacy law that grants California residents specific rights over their personal data, including the right to know what personal information is being collected, the right to request deletion, the right to opt out of data sales, and the right to non-discrimination for exercising these rights.

Continuing Education credits earned by completing credentials, attending events, or maintaining professional activity within the COMPEL ecosystem.

Continuing Education credits are the unit of measurement for ongoing professional development required to maintain COMPEL certifications.

The AI Center of Excellence is the operational nucleus of AI transformation -- the organizational structure that transforms individual AI capability into enterprise AI capacity.

A certification body is an organization authorized to assess and formally certify that individuals, systems, products, or organizations meet the requirements defined by specific standards or qualification frameworks.

Change architecture is the deliberate, comprehensive design of how organizational change will be structured, sequenced, resourced, and governed across an enterprise-scale AI transformation.

Change capacity management is the assessment, monitoring, and deliberate management of how much organizational change the workforce and leadership can absorb at any given time without experiencing change fatigue, disengagement, or active resistance.

Change management is the structured approach to transitioning individuals, teams, and organizations from a current state to a desired future state.

A change network is a distributed group of change advocates and champions embedded across the organization's departments and levels who support AI transformation by communicating key messages, coaching colleagues through transitions, surfacing feedback and concerns from the front lines, and modeling new behaviors in their daily work.

Change saturation is the practical limit on how much simultaneous organizational change a team, department, or enterprise can absorb effectively.

Chaos engineering is the discipline of deliberately introducing controlled failures, disruptions, and adverse conditions into a system's production or staging environment to test its resilience and discover weaknesses before they cause real incidents.

A chargeback model is a financial governance mechanism where business units are billed for their actual consumption of shared AI services, infrastructure, compute resources, and platform capabilities, creating cost transparency and incentivizing efficient resource use.

The Chief AI Officer (CAIO) is an emerging C-suite role responsible for an organization's overall AI strategy, governance, and transformation program.

The Chief Data Officer is a C-suite executive responsible for enterprise data strategy, data governance, data quality, and data infrastructure.

Churn prediction is an AI application that predicts which customers are likely to stop using a product or service within a defined timeframe, enabling proactive retention efforts before the customer leaves.

A CI/CD (Continuous Integration/Continuous Deployment) pipeline is an automated workflow that builds, tests, and deploys software changes through a series of stages, catching errors early and enabling rapid, reliable releases.

A circuit breaker is a resilience design pattern that automatically stops an AI system from sending requests to a failing downstream service or component when it detects a pattern of errors, preventing cascading failures and giving the failing component time to recover.

Classification is a supervised learning task that assigns inputs to discrete categories.

Client discovery is the initial investigative phase of a COMPEL engagement where the practitioner gathers comprehensive information about the prospective client's business context, AI aspirations, organizational constraints, stakeholder landscape, budget parameters, previous transformation attempts, and current technology environment.

Cloud computing is the delivery of computing services -- servers, storage, processing power, databases, networking, and software -- over the internet on a pay-as-you-go basis rather than owning and maintaining physical infrastructure.

Cloud-native architecture refers to systems designed specifically to leverage cloud computing capabilities such as elastic scaling, distributed processing, managed services, containerized deployment, and microservice decomposition.

Clustering is an unsupervised learning technique that groups similar data points together based on shared characteristics, without requiring pre-labeled categories.

A co-development agreement is a contractual arrangement where two or more parties jointly develop AI capabilities, clearly specifying how intellectual property ownership, development costs, risks, access rights, and commercial benefits are shared between the parties.

Coalition analysis is a stakeholder management technique that maps the formal and informal alliances, power relationships, shared interests, and competing agendas among individuals and groups within an organization to understand who can be brought together to support transformation and who may collectively resist it.

COBIT (Control Objectives for Information and Related Technologies) is an IT governance and management framework developed by ISACA that provides a comprehensive set of controls, processes, and metrics for governing enterprise information and technology.

The formal governance document that establishes the AI Center of Excellence — defining its mandate, scope, membership, decision rights, funding model, service catalogue, and success metrics.

Cognitive load management is the deliberate practice of controlling the mental effort required for learning, comprehension, and task performance, ensuring that training materials, communications, and governance processes do not overwhelm participants with excessive complexity or information volume.

A community of practice (CoP) is a group of people who share a professional interest, domain, or challenge and interact regularly to deepen their knowledge, share experiences, solve problems, and develop their expertise collectively.

A COMPEL cycle is a single iteration through all six stages (Calibrate, Organize, Model, Produce, Evaluate, Learn), typically lasting 12 weeks with a contextual range of 8 to 16 weeks.

The COMPEL Engagement Lifecycle is the five-phase structure for managing a COMPEL transformation consulting project, distinct from the six-stage COMPEL methodology itself.

The COMPEL Four Pillars are the fundamental organizing dimensions of AI transformation in the COMPEL framework: People (culture, skills, change management, leadership), Process (workflows, operations, governance procedures, service management), Technology (infrastructure, platforms, data architecture, AI models, security), and Governance (policies, compliance, ethics, risk management, oversight).

COMPEL is a structured, iterative six-stage methodology for enterprise AI transformation, standing for Calibrate, Organize, Model, Produce, Evaluate, Learn.

The COMPEL Lifecycle is the six-stage transformation methodology that structures all AI transformation work: Calibrate (assess current maturity), Organize (align stakeholders and form teams), Model (design the target state and roadmap), Produce (execute the transformation plan), Evaluate (measure outcomes and assess progress), and Learn (capture lessons and feed insights into the next cycle).

A low-to-moderate rigor credential requiring 6-8 hours that proves ability to perform a specific transformation task (e.g., readiness assessment, regulatory mapping, LLM governance).

Competency-based assessment is an evaluation approach that measures whether a person can demonstrate specific professional skills and knowledge in realistic practice contexts, rather than testing theoretical knowledge through traditional examinations alone.

A competitive moat is a durable competitive advantage that is exceptionally difficult for rivals to replicate.

Compliance posture refers to an organization's overall state of readiness and demonstrated adherence to the laws, regulations, standards, and internal policies applicable to its AI systems and data practices.

A compute budget is the allocated financial and resource limit for AI workloads including model training, experimentation, inference processing, and data pipeline operations.

Computer vision is a field of artificial intelligence that enables computers to interpret and understand visual information from images, videos, and real-time camera feeds, replicating aspects of human visual perception through mathematical models.

Concept drift occurs when the underlying relationship between input data and the outcome being predicted changes over time.

Conformity assessment is a formal evaluation process that determines whether a product, system, organization, or process meets the specific requirements of a defined standard, regulation, or specification.

Consent management encompasses the technical systems, processes, and policies for collecting, recording, tracking, and honoring individuals' preferences regarding the use of their personal data, including data used to train, validate, or operate AI systems.

Consortium governance is the design and operation of governance structures for multi-organization AI collaborations where no single entity has unilateral authority and decisions must be reached through negotiation, voting, or consensus mechanisms among sovereign participants.

Constructivism is a learning theory positing that people actively build their understanding by connecting new information to their existing knowledge, experiences, and mental models, rather than passively absorbing information transmitted by an instructor.

Containerization is a technology that packages software applications and all their dependencies (libraries, configurations, runtime environments) into isolated, portable units called containers that run consistently across different computing environments.

A context window is the maximum amount of text (measured in tokens) that a large language model can process at one time.

Continuous improvement is the ongoing effort to enhance processes, capabilities, and outcomes through iterative learning and refinement.

A governed, prioritized backlog of improvement actions identified across all Evaluate and Learn stage reviews — covering control enhancements, workflow refinements, model updates, policy changes, and capability investments — that ensures organizational learning is translated into concrete, scheduled improvement work rather than good intentions.

A record that tracks the implementation and activation status of every governance control specified in the Control Requirements Matrix — confirming that each control has been built, tested, and is operational in the production environment.

A control framework is a structured, comprehensive set of policies, procedures, technical safeguards, and organizational measures designed to manage risks and ensure compliance within a specific domain.

A structured report that assesses the effectiveness of every active governance control — presenting evidence of control operation, pass/fail status against defined thresholds, exceptions, and remediation actions — to demonstrate that governance is working as designed rather than merely documented.

A comprehensive mapping of every governance control required for each AI system — specifying the control type (preventive, detective, corrective), the risk or policy it addresses, the evidence required to prove effectiveness, the owner, and the testing frequency.

A Convolutional Neural Network is a type of deep learning architecture designed specifically for processing visual data like images and videos.

Copyright is the legal protection granted to original creative works including text, images, music, software code, and other forms of expression.

A credential designation is an honorary professional title granted when a practitioner achieves a specific combination of credentials spanning multiple credential types within the COMPEL credential lattice.

The multi-dimensional credential architecture that extends beyond the traditional linear certification ladder to include micro-credentials, specializations, competency badges, joint credentials, and designations.

A personalized dashboard within the COMPEL platform that visualizes a practitioner's progress across the credential lattice.

Credential stacking is the practice of combining multiple smaller credentials (micro-credentials, competency badges) to build toward higher-level credentials (specializations, joint credentials) through formally defined accumulation rules.

Crisis management is the organized process of preparing for, responding to, recovering from, and learning from unexpected events that threaten an organization's AI transformation program, operations, reputation, or stakeholder relationships.

Cross-border data governance encompasses the policies, legal mechanisms, technical architectures, and organizational processes for managing data that flows between different countries, each with potentially different data protection laws, sovereignty requirements, and regulatory expectations.

A cross-domain diagnostic is an advanced COMPEL assessment technique that examines how capabilities in different maturity domains interact, influence each other, and create systemic patterns that are not visible when each domain is assessed in isolation.

Cross-functional collaboration is the practice of working across traditional organizational boundaries -- IT, business units, legal, finance, HR, compliance -- to achieve AI transformation objectives.

A cross-functional team brings together members from different organizational departments, disciplines, or specializations to work collaboratively toward a common objective, combining perspectives that no single function could provide alone.

Cross-organizational governance refers to the structures, policies, and decision-making processes that operate across organizational boundaries to enable coherent AI policy, consistent risk management, and aligned strategic direction among entities that do not share a single command hierarchy.

Cross-validation is a statistical technique for evaluating AI model performance by partitioning data into multiple subsets, systematically training the model on some subsets while testing on others, and averaging the results across all partitions.

Cultural transformation is the deliberate, sustained reshaping of an organization's values, beliefs, behaviors, and working practices to create an environment that supports AI adoption, data-driven decision-making, experimentation, and responsible innovation.

CRM software manages an organization's interactions with current and potential customers, tracking sales activities, customer communications, purchase history, and service interactions.

Data architecture is the design of how data is collected, ingested, stored, organized, integrated, transformed, governed, and made available across an enterprise to support AI capabilities, analytics, and business operations.

A data catalog is a searchable, organized inventory of all data assets within an organization, providing metadata about each dataset's location, format, schema, ownership, quality metrics, access permissions, lineage, and permitted uses.

Data classification is the process of categorizing data based on its sensitivity level, regulatory requirements, and business criticality into tiers such as public, internal, confidential, and restricted.

Data drift occurs when the statistical properties of the input data a deployed model receives change compared to the data it was trained on.

A data engineer is a professional responsible for building and maintaining the data infrastructure and pipelines that collect, store, transform, and deliver data to AI models and analytics consumers.

A data fabric is an architectural approach that provides a unified, intelligent data management layer across diverse and distributed data sources, environments, and formats, using automation, metadata, and AI-driven data management to make data accessible and usable regardless of where it physically resides.

Data governance encompasses the organizational processes, policies, standards, and accountability structures that ensure data is accurate, consistent, secure, and used appropriately across the enterprise.

A system-specific plan that defines data lineage, quality standards, access controls, retention policies, consent requirements, and bias monitoring obligations for every dataset used by an AI system.

A data lake is a centralized storage repository that ingests and holds large volumes of raw data in its original format, whether structured, semi-structured, or unstructured, until it is needed for analysis, reporting, or AI model training.

A data lakehouse is a modern data architecture that combines the flexibility and scale of a data lake with the management features, performance, and data governance capabilities of a traditional data warehouse.

Data lineage is the documented, traceable history of a piece of data as it moves through an organization's systems, recording where it originated, how it was collected, what transformations were applied, where it was stored, who accessed it, and how it was ultimately used in AI models or business processes.

Data mesh is a decentralized data architecture and organizational approach where individual business domain teams own, produce, and maintain their data as discoverable, trustworthy data products, rather than centralizing all data management in a single data engineering team.

Data minimization is a core data protection principle, mandated by GDPR and adopted by many other privacy frameworks, requiring that organizations collect and retain only the personal data that is strictly necessary for a specific, stated purpose.

A data pipeline is an automated, orchestrated sequence of steps that moves data from source systems through extraction, transformation, validation, and loading processes to its destination, which may be a data warehouse, feature store, or directly an AI model's training or inference system.

Data poisoning is a type of attack where an adversary deliberately corrupts the data used to train an AI model, causing the model to learn incorrect patterns or behave in unintended ways.

A DPIA is a formal GDPR-required assessment when data processing poses high risk to individuals, evaluating necessity, proportionality, risks, and mitigations.

Data quality is the degree to which data meets requirements for accuracy, completeness, consistency, timeliness, validity, and uniqueness.

Data readiness is an assessment of whether the data required for an AI initiative is available, of sufficient quality, properly governed, legally accessible, and representative of the populations the AI system will serve.

A data readiness assessment is a structured evaluation of an organization's data ecosystem to determine its fitness for AI transformation, covering data quality metrics, data lineage documentation, metadata management maturity, access governance policies, storage and compute infrastructure scalability, and data team capabilities.

A data scientist is a professional who uses statistical analysis, machine learning, and programming to extract insights from data and build predictive or generative models.

A data steward is an individual formally responsible for the quality, governance, and appropriate use of data within a specific domain or business function.

A decision log is a formal, maintained record of significant decisions made during an AI transformation program, documenting the context and problem statement, alternatives considered, decision criteria, the decision reached, the rationale for the choice, who made the decision, what authority they had, and the expected consequences.

Decision provenance is the complete, traceable record of how an AI decision was reached, encompassing the input data, model version, algorithm parameters, intermediate reasoning steps, tool calls, and contextual factors that contributed to a specific output.

Decision rights are formally documented authorities specifying who can approve what within the AI transformation program.

Deep learning is a subset of machine learning that uses artificial neural networks with many layers (hence 'deep') to automatically learn complex patterns and representations from large amounts of data.

Defense in depth is a security strategy that implements multiple, layered defensive mechanisms throughout an AI system so that if any single layer is breached, other layers continue to provide protection.

A delegation framework is a governance structure that precisely defines what decisions, actions, and resource commitments an AI agent is authorized to make independently, what requires human approval before proceeding, what escalation paths exist for edge cases, and what hard boundaries the agent must never cross.

Demand forecasting uses AI to predict future customer demand for products or services, enabling optimized inventory management, production planning, workforce scheduling, and supply chain operations.

Demographic parity is a mathematical fairness criterion requiring that an AI system's positive outcomes (such as loan approvals, job interview invitations, or benefit eligibility) are distributed equally across different demographic groups, regardless of the group's representation in the underlying data.

A comprehensive, sign-off-based checklist that verifies all technical, governance, operational, and organizational prerequisites have been met before an AI system is deployed to production.

A pinnacle-tier credential recognizing cross-stack mastery across multiple credential types in the COMPEL ecosystem.

DevOps is a set of cultural practices, processes, and tools that integrate software development (Dev) and IT operations (Ops) to enable organizations to deliver software changes more frequently, reliably, and with higher quality.

Differential privacy is a rigorous mathematical framework for sharing data, statistical analyses, or machine learning model outputs while providing formal guarantees that no individual's private information can be inferred from the results.

Dimensionality reduction is a technique that simplifies complex datasets with many variables by identifying the most important underlying factors and representing the data in fewer dimensions.

Disaster recovery encompasses the plans, processes, and technical infrastructure for restoring AI systems, data, and services after a catastrophic failure such as data center outages, major security breaches, data corruption, or natural disasters.

Discriminative AI models analyze input data to classify it, predict outcomes, or identify patterns.

Disparate impact occurs when an AI system's decisions disproportionately and negatively affect a particular demographic group even though the system does not explicitly use protected characteristics such as race, gender, or age as input variables.

DMAIC (Define, Measure, Analyze, Improve, Control) is the five-phase improvement cycle from Lean Six Sigma methodology.

Drift detection is automated monitoring that identifies when the statistical properties of input data or model outputs have shifted significantly from baseline measurements established during model training or initial deployment.

Due diligence is the comprehensive investigation and risk evaluation of an organization, technology, vendor, or partnership opportunity conducted before making a significant commitment such as an acquisition, major vendor contract, or strategic partnership.

Ecosystem strategy designs external partnerships, collaborations, and alliances providing capabilities the organization cannot build alone.

Edge computing is the practice of processing data near its source (at the 'edge' of the network) rather than sending all data to a centralized cloud data center, enabling low-latency AI inference, reduced bandwidth consumption, and operation in environments with limited or intermittent connectivity.

Edge deployment refers to running AI models on devices located close to where data is generated -- factory equipment, IoT sensors, retail stores, or branch offices -- rather than in a centralized cloud.

An embedding is a mathematical representation that converts text, images, or other complex data into dense numerical vectors (lists of numbers) that capture semantic meaning and relationships.

Engagement architecture is the comprehensive design of a COMPEL consulting engagement, encompassing its scope (which domains and pillars are included), phases (how the work is sequenced), workstreams (how parallel activities are organized), deliverables (what tangible outputs will be produced), timeline (how long the engagement will run), team composition (what roles and skills are needed), governance structure (how decisions will be made and progress tracked), and commercial model (how the work is priced and paid for).

The Enterprise AI Maturity Spectrum defines five levels of organizational AI capability: Level 1 (Foundational -- scattered, ungoverned experimentation), Level 2 (Developing -- intentional investment with initial governance), Level 3 (Defined -- repeatable, standardized AI delivery), Level 4 (Advanced -- AI embedded in core operations with proactive governance), and Level 5 (Transformational -- AI reshapes the business model).

Enterprise AI transformation is the coordinated, organization-wide effort to embed artificial intelligence into the strategic fabric of a large or complex organization — spanning multiple business units, geographies, regulatory jurisdictions, and technology stacks.

ERP systems are integrated business software suites that manage core organizational processes including finance, supply chain, manufacturing, human resources, and procurement.

Enterprise Transformation Architecture is a comprehensive, integrated blueprint that unifies AI strategy, organizational design, technology architecture, governance frameworks, change management, and measurement systems into a coherent whole for enterprise-scale AI transformation.

Equalized odds is a mathematical fairness criterion requiring that an AI system has equal true positive rates and equal false positive rates across different demographic groups, meaning the system is equally accurate for each group and distributes its errors fairly.

ESG is a framework for evaluating corporate behavior and sustainability across three dimensions: Environmental (climate impact, resource usage), Social (labor practices, community impact, diversity), and Governance (corporate ethics, board oversight, transparency).

An Ethical Impact Assessment is a mandatory evaluation conducted before any AI system moves from development to production, systematically assessing potential harms, affected populations, and mitigation strategies.

Ethics by design is the approach of integrating ethical considerations into every stage of the AI development lifecycle rather than reviewing ethics after the system is built.

An ethics review process evaluates proposed AI projects for ethical implications before authorization, defining triggers, criteria, review body, decision options, and appeals.

An ETL (Extract-Transform-Load) or ELT (Extract-Load-Transform) pipeline is a data processing workflow that moves data from source systems into target repositories where it can be used for AI training and operations.

The EU AI Act (Regulation 2024/1689) is the world's first comprehensive legal framework for regulating artificial intelligence, adopted by the European Parliament in March 2024 and entering into force in August 2024.

Evaluate is the fifth COMPEL stage, focused on rigorously measuring what the current cycle achieved against planned objectives.

The Evaluate stage is the fifth stage of the COMPEL lifecycle where the outcomes of the transformation program are systematically measured against the objectives established during the Model stage, maturity progression is re-assessed using the 18-domain model, stakeholder satisfaction is gauged, and the overall effectiveness and efficiency of the program are critically examined.

An evidence chain is a sequence of related governance artifacts that together tell a complete, traceable story from strategic intent through operational implementation.

The documented configuration of all evidence collection processes — defining what evidence is gathered, from which systems, on what schedule, in what format, and how it is stored and linked to governance controls — so that the Evidence Pack can be assembled continuously and automatically rather than scrambled together before audit.

The complete, auditable collection of artifacts, test results, decision records, and attestations that demonstrate an AI system meets its governance, compliance, and operational requirements.

Executive coaching in the COMPEL context is the structured, one-on-one developmental relationship where an AITGP-level consultant helps senior leaders develop the mindset, capabilities, and behaviors needed to champion, sustain, and personally embody AI transformation in their organizations.

The Executive Sponsor is the C-suite champion who provides strategic direction, budget authority, and organizational air cover for AI transformation.

Executive sponsorship is the active, visible, sustained commitment from a senior organizational leader who champions the AI transformation program by securing funding, allocating resources, removing organizational barriers, resolving cross-functional conflicts, communicating the transformation vision, and holding the organization accountable for progress.

Experiential learning is an educational approach grounded in the theory that lasting knowledge and skill development come from direct experience followed by structured reflection, conceptualization, and active experimentation.

Explainability is the degree to which an AI system's decision-making process can be understood and communicated to humans.

A documented specification of the minimum explainability standard each AI system must meet — defining what must be explainable, to whom, in what form, and within what timeframe — aligned to the system's risk classification, regulatory context, and stakeholder needs.

Explainable AI (XAI) is a field of research and practice focused on developing techniques, tools, and methodologies that make AI decision-making processes understandable to humans.

A formal mapping between a credential issued by an external organization (such as a recognized training partner, AWS, Google, or Microsoft) and recognition within the COMPEL credential ecosystem.

A structured mapping between recognized technical training program completions and COMPEL credential ecosystem entry points.

The F1 score is a model performance metric that combines precision and recall into a single balanced measure, calculated as the harmonic mean of the two.

Facilitation is the professional skill of guiding group discussions, workshops, and collaborative sessions to achieve productive outcomes by managing group dynamics, encouraging diverse participation, maintaining focus on objectives, and synthesizing collective input without imposing the facilitator's own views or conclusions.

Fairness in AI is the principle that AI systems should produce equitable outcomes across different demographic groups and not perpetuate or amplify existing societal biases.

Fairness engineering is the technical discipline of detecting and mitigating bias in AI systems through systematic processes applied throughout the model lifecycle.

A feature store is a centralized, managed repository for storing, versioning, and serving the processed data features (engineered variables) used to train and run AI models, enabling feature reuse across teams, ensuring consistency between training and serving environments, and reducing the redundant data processing that occurs when each team independently creates the same features.

Federated governance sets central AI standards while giving business units implementation autonomy within defined boundaries.

Federated learning is a machine learning approach where a model is trained across multiple devices, servers, or organizations holding local data, without exchanging the raw data itself.

In organizational design, a federated model distributes AI capability across business units while maintaining a central team that provides standards, shared infrastructure, coordination, and governance.

A federated organizational model distributes AI capability across business units with central coordination, balancing local autonomy with enterprise consistency.

A feedback loop in AI occurs when an AI system's outputs influence its future inputs, creating a self-reinforcing cycle that can either improve or degrade performance over time.

Fine-tuning is the process of further training a pre-trained AI model on a specific dataset to adapt it for a particular task or domain.

FinOps (Financial Operations) is the practice of bringing financial accountability, transparency, and optimization to variable cloud and infrastructure spending through real-time cost visibility, collaborative decision-making between engineering and finance teams, and continuous optimization of resource utilization.

A foundation model is a large pre-trained AI model that serves as a base for multiple downstream applications.

The Four Pillars -- People, Process, Technology, and Governance -- are the four interdependent structural foundations of AI transformation in the COMPEL framework.

Framework interoperability is the ability of different management, governance, and delivery frameworks such as COMPEL, SAFe, TOGAF, ITIL, COBIT, PMBOK, and Lean Six Sigma to work together effectively within an organization without creating conflicting requirements, redundant processes, or framework fatigue among practitioners.

A full transformation engagement is a COMPEL consulting arrangement that spans the complete COMPEL lifecycle from Calibrate through Learn, typically running six to twenty-four months, involving cross-functional teams, and requiring sustained executive sponsorship.

Function calling is the capability of modern LLMs to produce structured calls to external tools and APIs as part of their output, enabling AI agents to interact with enterprise systems and take real-world actions.

Gap analysis is the systematic comparison of an organization's current state (as determined by maturity assessment) to its desired future state (as defined by strategic objectives), producing a detailed map of specific capability gaps that must be closed through targeted transformation initiatives.

The formal record of a stage gate review decision — documenting the inputs reviewed, the pass/fail outcome, the conditions or exceptions attached to the decision, the named decision-makers, and any required remediation actions before the next gate.

The General Data Protection Regulation (GDPR) is the European Union's comprehensive data protection law that governs how personal data of EU residents is collected, processed, stored, and transferred, imposing strict requirements for lawful basis, consent, data minimization, purpose limitation, individual rights (access, deletion, portability), data protection impact assessments, and breach notification.

Generative AI refers to artificial intelligence systems capable of creating new content, including text, images, code, music, video, and synthetic data, based on patterns learned from large training datasets.

A defined mechanism — preventive, detective, or corrective — that enforces policy compliance, mitigates identified risks, or ensures operational integrity for AI systems.

Governance harmonization is the deliberate process of aligning different AI governance frameworks, policies, standards, and practices across organizational units, business entities, jurisdictions, or partner organizations to create a coherent, non-contradictory governance environment that participants can comply with efficiently.

Governance maturity measures the sophistication of AI governance from ad hoc and reactive to optimized and continuously improving.

Governance theater is an anti-pattern where an organization builds the visible apparatus of AI governance -- policies, committees, ethics boards, review processes, published principles -- without operationalizing any of it.

Governance-as-enabler is a strategic design philosophy that positions AI governance not as a restrictive control mechanism that slows innovation but as an accelerant that enables the organization to move faster with confidence by providing clear guidelines, pre-approved pathways for common scenarios, and efficient review processes that reduce uncertainty and rework.

A GPU is a specialized processor originally designed for rendering graphics in video games, now widely repurposed for AI workloads.

Graceful degradation is the design principle and architectural capability that allows an AI system to continue operating at reduced functionality rather than failing completely when components break, resources become constrained, or performance degrades.

A GRC (Governance, Risk, and Compliance) platform is specialized software that automates governance workflows, approval tracking, evidence chain visualization, compliance reporting, and risk management processes.

Grounding refers to techniques that connect AI model outputs to factual, verifiable information sources rather than relying solely on patterns learned during training.

Guardrails are the safety boundaries, constraints, filters, and monitoring mechanisms built into AI systems to prevent harmful, inappropriate, unauthorized, or out-of-scope behaviors and outputs.

Hallucination is the phenomenon where an AI system, particularly a large language model, generates output that is plausible-sounding and confidently stated but factually incorrect, fabricated, or unsupported by the model's training data or any real-world source.

The Health Insurance Portability and Accountability Act (HIPAA) is a US federal law that establishes strict requirements for protecting sensitive patient health information (Protected Health Information, or PHI) from unauthorized disclosure, with severe penalties for violations.

A horizon portfolio allocates AI investments across three time horizons: near-term quick wins, medium-term capability building, and long-term strategic bets, ensuring continuous value while investing in future capabilities.

Human oversight in the context of AI governance refers to the organizational mechanisms, processes, and technical controls that ensure qualified humans maintain meaningful authority over AI system decisions throughout the system lifecycle.

A formal specification of every decision point in an AI system's operation where human review, approval, or override is required — including the trigger conditions, review timeline, escalation path, and documentation requirements for each rule.

Human-in-the-loop (HITL) is an AI system design pattern where a human must actively review, approve, and authorize each AI decision or action before it is executed, providing maximum human oversight at the cost of reduced speed and scalability.

Human-on-the-loop (HOTL) is an AI system design pattern where the AI makes and executes decisions autonomously, but a human monitors the process through dashboards and alerts and can intervene to override, pause, or adjust the system when anomalies or problems are detected.

A Hybrid Center of Excellence is an organizational model where a central team owns AI standards, governance, shared platforms, and complex cross-functional initiatives, while embedded AI teams within business units handle domain-specific delivery.

The Hype Cycle is a Gartner model describing the typical progression of emerging technologies through five phases: Technology Trigger (initial breakthrough generates interest), Peak of Inflated Expectations (publicity produces unrealistic enthusiasm), Trough of Disillusionment (implementations fail to deliver on hype), Slope of Enlightenment (practical benefits become understood), and Plateau of Productivity (mainstream adoption with realistic expectations).

In-context learning is the simplest form of AI agent adaptation, where the model adapts its behavior based on information in its current context window -- the conversation history, task instructions, retrieved documents, and tool outputs -- without changing its underlying model weights.

A structured review of all AI-related incidents, near-misses, and emerging risks that occurred during the evaluation period — including root cause analysis, control failure attribution, and required remediation actions — to ensure that the organization learns from operational experience and updates its risk profile accordingly.

AI incident response encompasses the defined procedures for investigating and remediating AI-related events such as model failures, bias discoveries, data breaches, safety incidents, or unexpected behavioral changes.

Indemnification is a contractual provision where one party compensates another for specified losses from AI system failures, data breaches, or IP infringement.

Inference is the process of using a trained AI model to make predictions or generate outputs on new, previously unseen data.

The Influence-Interest Matrix is a stakeholder analysis tool that maps individuals or groups along two dimensions: their level of influence over transformation outcomes and their level of interest in transformation activities.

Information asymmetry occurs when different teams possess different knowledge about project status or risks, leading to misaligned decisions and coordination failures.

Infrastructure as Code is the practice of managing and provisioning computing infrastructure through machine-readable configuration files rather than manual setup processes.

Initiative sequencing is the strategic ordering of transformation activities based on dependencies between initiatives, organizational readiness and absorption capacity, quick-win opportunity timing, resource availability, regulatory deadlines, and the compounding value that certain foundational capabilities provide to subsequent initiatives.

Internal audit is an independent assurance function evaluating risk management, governance, and controls governing AI.

Interpretability is the degree to which a human can understand the internal mechanisms and decision-making logic of an AI model, enabling meaningful inspection of how inputs are transformed into outputs.

An investment thesis documents why a specific AI investment will create value, what assumptions underlie returns, and how success will be measured.

ISO/IEC 27001 is the international standard for information security management systems (ISMS).

ISO/IEC 42001:2023 is the first international management system standard for artificial intelligence, published jointly by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC).

ITIL is a widely adopted framework for IT service management that defines processes for incident management, change management, service level management, configuration management, and continual service improvement.

The J-curve effect describes the common pattern in AI transformation where organizational performance initially dips before improving beyond the original level, forming a J-shaped curve when plotted over time.

Jailbreaking is the practice of crafting inputs, prompts, or interactions designed to manipulate an AI system into bypassing its built-in safety restrictions, content filters, or behavioral guidelines to produce prohibited content, reveal confidential information, or perform unauthorized actions.

Under data protection laws such as GDPR, a joint controller arrangement exists when two or more organizations jointly determine the purposes and means of processing personal data, creating shared legal responsibilities for data protection compliance.

A co-branded credential developed in partnership with a recognized training organization that requires both technical expertise and COMPEL transformation methodology competency.

A joint venture is a business arrangement where two or more organizations combine resources, expertise, and data to pursue a shared AI initiative while maintaining their separate organizational identities.

JSON (JavaScript Object Notation) is a lightweight, human-readable data format used extensively in AI systems for API communication, configuration files, model metadata, and structured data exchange between applications.

Judicial review is the process by which courts examine the legality, fairness, and procedural propriety of decisions made by government agencies, public bodies, or AI systems used in administrative or legal decision-making.

K-fold cross-validation is a model evaluation technique that provides a more reliable estimate of model performance than a single train-test split.

Kanban is a visual workflow management method that uses boards divided into columns representing stages of work, with cards representing individual tasks that move across the board from left to right as they progress.

A Key Performance Indicator is a quantifiable measurement used to evaluate how effectively an organization or initiative is achieving its objectives.

A Key Risk Indicator (KRI) is a measurable metric that provides early warning of increasing risk exposure before risks materialize as actual incidents or losses.

A kill switch is an immediate, unconditional mechanism to halt an AI agent's operation.

In COMPEL, the knowledge base is the persistent organizational repository of governance knowledge, best practices, lessons learned, reusable patterns, and cautionary tales that accumulates across transformation cycles.

A knowledge graph is a structured representation of real-world entities (people, places, concepts, products) and their relationships, stored in a graph database that enables sophisticated querying and reasoning.

Knowledge management is the organizational practice of capturing, organizing, sharing, and applying institutional knowledge to improve decision-making and performance over time.

Knowledge transfer is the deliberate process of transmitting expertise, skills, and understanding from external consultants to client team members, or from experienced practitioners to newer colleagues, ensuring the receiving organization retains the capability to sustain and continue transformation after the engagement ends.

A structured, cadenced report that measures actual AI system and program performance against the KPIs defined in the Value Thesis Register and Measurement Model — presenting actuals vs.

Kubernetes is an open-source container orchestration platform that automates the deployment, scaling, management, and networking of containerized applications across clusters of machines.

Labeling (also called annotation) is the process of tagging data with correct answers to create training datasets for supervised learning.

A Large Language Model is a massive AI model -- typically based on the transformer architecture and containing billions to trillions of parameters -- trained on enormous amounts of text data to understand and generate human language.

Latency is the time delay between sending a request to an AI system and receiving a response, typically measured in milliseconds.

Leadership transition management maintains AI transformation momentum when key leaders change due to promotion, departure, or reorganization.

A set of investment policies that govern how funding flows to AI value streams without requiring per-project business cases — typically covering portfolio participation rules, capacity allocation by horizon, approval thresholds for epics, and continuous business owner engagement.

Lean Six Sigma is a process improvement methodology that combines Lean principles (eliminating waste, maximizing value) with Six Sigma statistical techniques (reducing variation, achieving consistent quality).

Learn is the sixth and final COMPEL stage, and paradoxically both the most strategically undervalued and most organizationally consequential.

The Learn stage is the sixth and final stage of the COMPEL lifecycle where the organization systematically captures lessons learned, codifies new knowledge gained during transformation, shares insights across the enterprise, and feeds those learnings back into the beginning of the next cycle.

A learning organization, as conceptualized by Peter Senge, is an enterprise that continuously transforms itself through the expansion of its capacity to learn.

Least privilege is a foundational security principle requiring that AI agents receive access only to the minimum set of tools, data, and system permissions necessary to perform their defined function.

Load balancing distributes incoming requests across multiple servers or model instances to prevent overload, ensuring consistent performance and high availability.

Machine Learning is a subset of AI where systems learn patterns from data rather than being explicitly programmed with rules.

Master Data Management is the set of processes, governance structures, and technology for ensuring consistent, authoritative definitions of key business entities -- customers, products, suppliers, locations, employees -- across the entire enterprise.

A maturity assessment is a structured, evidence-based evaluation that measures an organization's capabilities, practices, and governance against a defined maturity model, producing numerical scores and qualitative findings that indicate current state, identify gaps, and guide improvement priorities.

A scored assessment of the organization's current AI capability maturity across people, process, technology, and governance dimensions, establishing the starting point against which transformation progress is measured.

The maturity plateau is a COMPEL-identified anti-pattern where organizations make genuine early progress in AI transformation -- achieving production deployments, measurable business impact, and functioning governance -- but then stall at an intermediate maturity level (typically Level 2-3), unable to advance further.

A maturity progression dashboard is a visual governance tool that tracks an organization's advancement across the 18 COMPEL maturity domains over time, displaying current scores, historical trends, targets, and the gaps remaining for each domain.

The structured framework for quantifying AI transformation progress and outcomes across four levels: strategic KPIs (organization-level), portfolio KPIs (aggregate across use cases), use-case KPIs (individual initiative performance), and operational KPIs (system-level health).

Memory poisoning is an attack targeting AI agents with persistent memory, where an adversary manipulates what the agent remembers to permanently alter its behavior across future sessions.

Metadata is data that describes other data -- information about a dataset's source, format, creation date, quality metrics, ownership, access permissions, update frequency, and usage history.

Methodology benchmarking is the systematic comparison of AI transformation methodologies across different frameworks, practitioners, and organizations to identify best practices, performance standards, areas for improvement, and opportunities for innovation.

A focused, stackable credential requiring 12-20 hours of study with automated quiz and portfolio artifact assessment.

Microservices is an architectural pattern where applications are built as a collection of small, independent services that communicate through well-defined APIs, each responsible for a specific function and deployable independently.

An ML engineer is a professional who specializes in building production-quality machine learning systems, bridging the gap between data science (model development) and software engineering (production deployment).

MLOps (Machine Learning Operations) is the set of practices, tools, and cultural patterns that enable organizations to deploy, monitor, and maintain machine learning models in production reliably and at scale.

In AI and machine learning, a model is a mathematical representation learned from data that can make predictions or generate outputs.

Model is the third COMPEL stage, where assessment findings and organizational readiness converge into a concrete, evidence-based transformation plan for the current 12-week cycle.

A model card is a standardized documentation template that describes an AI model's intended use, training data, performance characteristics across different populations, known limitations, fairness evaluations, and ethical considerations.

Model drift is the degradation of an AI model's performance over time caused by changes in the statistical properties of the input data, the target variable, or the relationship between them.

Model lifecycle management is the governance discipline of maintaining visibility, control, and accountability over AI models from initial conception through production deployment, monitoring, retraining, and eventual retirement.

Model monitoring is the continuous, automated observation of AI models operating in production to track performance metrics (accuracy, latency, throughput), detect data drift and concept drift, identify anomalous behavior, monitor fairness metrics, and ensure the model continues to operate within the acceptable parameters defined by its governance framework.

A model registry is a centralized, versioned repository for storing, cataloging, and managing AI models throughout their lifecycle, maintaining metadata about each model's training data, hyperparameters, performance metrics, deployment status, owner, and governance approval status.

Model risk is the risk of adverse consequences arising from errors, limitations, or inappropriate use of AI models.

Model Risk Management is a governance discipline originating in financial services that provides structured approaches to validating, documenting, monitoring, and governing AI/ML models used in decision-making.

The Model stage is the third COMPEL lifecycle stage designing the target state based on Calibrate gaps and Organize priorities, producing the transformation roadmap, governance framework, technology blueprint, and measurement framework.

Model validation is the independent assessment of an AI model's performance, fairness, robustness, and compliance before it is deployed to production.

A model validation pipeline automates quality checks, tests, fairness assessments, and security scans that models must pass before production authorization.

A multi-agent system (MAS) is an AI architecture in which multiple autonomous agents, each with specialized capabilities or knowledge domains, collaborate to accomplish tasks that no single agent could handle effectively alone.

Multi-modal AI refers to AI systems that can process and reason across multiple types of data simultaneously, such as text, images, audio, and video.

Multi-workstream coordination is the discipline of keeping parallel transformation activities across the People, Process, Technology, and Governance pillars aligned and progressing in concert during the Produce stage.

Multinational governance architecture designs AI governance operating across multiple countries, balancing global consistency with local regulatory and cultural adaptation.

Natural Language Processing is a branch of AI focused on enabling computers to understand, interpret, and generate human language.

Net Present Value is a financial calculation that determines the current value of all future cash flows from an AI investment minus the initial cost, using a discount rate that reflects the time value of money and investment risk.

A network effect occurs when an AI system or platform becomes more valuable as more people or data flows through it, creating a self-reinforcing cycle: better models attract more users, more users generate more data, more data trains better models.

An artificial neural network is a computing system loosely inspired by the human brain, consisting of layers of interconnected nodes (neurons) that process data by adjusting numerical weights during training.

The NIST AI Risk Management Framework (AI RMF 1.0), published by the National Institute of Standards and Technology in January 2023, is a voluntary framework for managing risks associated with the design, development, deployment, and evaluation of AI products and services.

An NDA establishes confidentiality obligations between parties, protecting sensitive information during AI engagements, vendor evaluations, and partnerships.

Observability is the comprehensive ability to understand the internal state, behavior, and health of an AI system by examining its external outputs, including logs, metrics, traces, and events.

An operating model defines how an enterprise structures its teams, processes, governance mechanisms, and technology to deliver its strategy.

Operating model design is the process of defining how an organization's AI capabilities will be structurally organized, staffed, funded, governed, and operated to deliver sustainable value at enterprise scale.

The assessed capability of an organization to sustain AI operations across 10 interdependent dimensions: strategy alignment, governance maturity, operating model, workforce capability, data readiness, technology infrastructure, monitoring and observability, vendor dependency management, compliance readiness, and change and adoption.

Operational resilience is the ability of an organization to prevent, prepare for, respond to, recover from, and learn from operational disruptions to its AI systems and AI-dependent business processes.

Opportunity cost is the potential value lost by choosing one AI initiative over another when resources are limited.

An oral defense is a live examination where COMPEL certification candidates at Levels 3 and 4 present and defend their capstone work before a panel of experienced evaluators who probe the depth of understanding, professional judgment, and integrated mastery demonstrated in the submission.

Organizational readiness for AI transformation is the degree to which an organization's people, culture, processes, and structures are prepared to adopt and benefit from AI.

Organize is the second COMPEL stage, focused on building the human and structural infrastructure required to execute transformation.

The Organize stage is the second COMPEL lifecycle stage translating Calibrate findings into mobilization through stakeholder alignment, team formation, governance establishment, and resource securing.

Overfitting occurs when an AI model learns the training data too precisely -- memorizing specific examples including their noise and anomalies rather than learning generalizable patterns -- and consequently performs poorly on new, unseen data.

A parameter is a learned numerical value within an AI model that is adjusted during training to improve the model's ability to make accurate predictions.

A structured update to the organization's AI pattern library — capturing reusable design patterns, anti-patterns, successful control configurations, and workflow redesign templates discovered during the current transformation cycle — so that future use cases benefit from accumulated organizational experience.

PCI DSS (Payment Card Industry Data Security Standard) is a set of security standards for organizations that handle, process, or store credit card information.

Peer review assessment is a structured evaluation process in which capstone project submissions are reviewed by qualified peers using standardized rubrics with weighted criteria, defined scoring levels, minimum passing thresholds, and required reviewer counts.

A structured evaluation framework used by qualified reviewers to assess specialization capstone projects and joint credential defenses.

Penetration testing is the authorized, controlled simulation of real-world attacks against an AI system to identify exploitable security vulnerabilities before malicious actors can discover and exploit them.

Persistent memory extends an AI agent's learning beyond a single session by storing information -- facts, preferences, outcomes, strategies -- in an external memory system that is retrieved when processing new tasks.

A pilot program is a controlled, limited-scope initial deployment of an AI solution in a real operational environment, designed to test feasibility, measure actual impact, identify integration challenges, gather user feedback, and validate the business case before committing to full-scale enterprise implementation.

Pilot purgatory is a COMPEL-identified anti-pattern where organizations launch numerous AI pilot projects but never build the governance, data infrastructure, organizational capability, or production readiness to move them beyond the pilot stage.

The pilot-to-production gap describes the common phenomenon where AI proofs of concept demonstrate impressive results in controlled environments but never scale to full production deployment.

A playbook is a documented set of step-by-step procedures, decision trees, and communication templates for handling specific operational scenarios such as AI incident response, model deployment, stakeholder escalation, or governance review.

A structured inventory and gap analysis of all existing organizational policies relevant to AI — covering data, security, privacy, ethics, procurement, and acceptable use — that identifies which policies need to be created, updated, or retired to support the AI transformation program.

Policy lifecycle management covers creating, reviewing, approving, implementing, monitoring, updating, and retiring AI governance policies ensuring they remain current.

A tracked register of all policy changes, additions, and retirements triggered by insights from the Evaluate stage — documenting the rationale for each policy change, the approver, the effective date, and the affected systems or processes.

Political navigation is the professional skill of understanding and working within organizational power dynamics, informal influence networks, competing agendas, and institutional politics to advance AI transformation objectives.

Portfolio management in the COMPEL context is the centralized governance and optimization of a collection of AI transformation programs, projects, and operational activities managed together to achieve enterprise strategic objectives.

Portfolio risk aggregation combines individual program risks into portfolio-level views revealing systemic patterns, correlated exposures, and concentration risks invisible at program level.

Thematic investment categories defined at the portfolio level that translate enterprise strategy into the differentiated business outcomes the AI portfolio must deliver.

A 3-5 year directional statement of the AI portfolio target state, describing the capabilities, business outcomes, and operating posture the organization intends to reach.

A post-mortem is a structured review conducted after an AI incident, project milestone, or engagement conclusion to analyze what happened, understand root causes, identify lessons learned, and develop actionable improvements for the future.

Precision is a model performance metric measuring the proportion of positive predictions that are actually correct -- in other words, when the model says 'yes,' how often is it right? High precision means few false positives (false alarms).

Predictive maintenance uses AI to predict when equipment will fail so maintenance can be performed just before failure occurs, rather than on a fixed schedule (preventive maintenance) or after failure (reactive maintenance).

PRINCE2 (Projects in Controlled Environments) is a structured project management methodology widely used in government, regulated industries, and large enterprises.

Privacy in the AI context goes beyond compliance with regulations like GDPR or CCPA to encompass a broader commitment to responsible data stewardship.

Privacy by design embeds data privacy protections into AI system design from the earliest stages rather than adding controls afterward.

Proactive regulatory engagement is the strategic practice of actively participating in regulatory development processes rather than passively waiting for final rules and then scrambling to comply.

Produce is the fourth COMPEL stage and the execution engine of each 12-week cycle.

The Produce stage is the fourth COMPEL lifecycle stage executing the transformation plan across all four pillars simultaneously including technology deployment, model development, process redesign, governance operationalization, and change management.

The verified state in which an AI system meets all prerequisites for safe, governed operation in a production environment.

Prompt engineering is the practice of designing and refining the text inputs (prompts) given to a large language model to produce desired outputs.

Prompt injection is a security attack where malicious instructions are hidden in input data to manipulate an AI agent's behavior, potentially causing it to ignore safety guidelines, reveal sensitive information, or take unauthorized actions.

A Proof of Concept is a small-scale implementation that demonstrates the feasibility of an AI solution in a controlled environment, typically using sample data and simplified conditions.

A provenance graph represents the complete chain of data sources, processing steps, and agent interactions leading to an AI output.

Pseudonymization replaces identifying information with artificial identifiers while maintaining a secured re-identification mapping.

Psychological safety is the shared belief within a team or organization that individuals can take interpersonal risks -- asking questions, admitting mistakes, proposing unconventional ideas, reporting problems -- without fear of punishment or ridicule.

Purpose limitation is the privacy principle ensuring that data collected for one purpose is not repurposed for AI training or other uses without appropriate consent, legal basis, and governance review.

Quality assurance for AI extends traditional software testing with model-specific validation processes to ensure AI systems meet defined standards for performance, reliability, fairness, and governance compliance.

A quality gate is a predefined checkpoint in a development or transformation process where deliverables must meet explicit quality criteria before work can proceed to the next stage.

Quantitative risk assessment is an approach to evaluating AI risks that uses numerical data, statistical methods, and mathematical models to estimate the probability and potential financial or operational impact of identified risks.

Quantization is an optimization technique that reduces the computational resources required to run an AI model by decreasing the numerical precision of its internal calculations, typically from 32-bit floating point to 16-bit, 8-bit, or even 4-bit representations.

Query optimization is the process of improving the efficiency of data retrieval operations to reduce latency (response time) and resource consumption (compute and storage costs).

A quick win is a transformation initiative strategically selected and designed to deliver visible, measurable, and broadly recognized value within a short timeframe (typically six to twelve weeks), building organizational momentum, stakeholder confidence, and political support for the broader, longer-term transformation program.

A role-accountability matrix that assigns Responsible, Accountable, Consulted, and Informed designations for every key AI governance decision, control activity, and operational process across the organization.

Retrieval-Augmented Generation (RAG) is an AI architecture pattern that enhances the accuracy and reliability of large language model outputs by first retrieving relevant information from external knowledge sources (databases, documents, knowledge bases) and then including that retrieved information in the context provided to the model for response generation.

A point-in-time, evidence-backed assessment of the organization's readiness to advance from one COMPEL stage to the next, scored across all 10 operational readiness dimensions with explicit pass/fail thresholds and remediation actions for any dimension below threshold.

Real-time processing involves generating AI predictions as events occur, typically delivering results within milliseconds to seconds.

Recall is a model performance metric measuring the proportion of actual positive cases that the model correctly identifies -- in other words, of all the real positives, how many did the model catch? High recall means few false negatives (missed cases).

A recommendation engine is an AI system that suggests relevant items -- products, content, actions, or connections -- to users based on their behavior, preferences, and similarities to other users.

Red teaming is a security and safety testing practice where a dedicated team deliberately attempts to find vulnerabilities, trigger unsafe behavior, or exploit weaknesses in an AI system.

Regression is a supervised learning task that predicts a continuous numerical value rather than a discrete category.

Regulatory compliance for AI encompasses the organizational processes and practices that ensure AI systems meet the requirements of applicable laws, regulations, and industry standards across all relevant jurisdictions.

Regulatory intelligence is the systematic, ongoing monitoring, analysis, and interpretation of regulatory developments, enforcement actions, policy proposals, guidance documents, and judicial decisions relevant to AI across all jurisdictions where an organization operates.

A regulatory sandbox is a controlled, supervised environment established by a regulatory authority where organizations can test innovative AI applications under relaxed or modified regulatory requirements, with close regulator oversight and structured learning objectives.

Reinforcement Learning (RL) is a machine learning paradigm where an agent learns by interacting with an environment and receiving rewards or penalties for its actions.

RLHF is the technique used to align large language model behavior with human preferences and safety requirements.

Resilience is the multidimensional capability of an AI system, transformation program, or organization to anticipate, withstand, respond to, recover from, and adapt to adverse events, disruptions, and changing conditions.

Responsible AI is the practice of designing, developing, and deploying AI systems in ways that are ethical, transparent, fair, accountable, and safe — and that actively avoid creating harm to individuals, groups, or society.

A formal record of the decision to retire or fundamentally redesign an AI use case — documenting the evidence that triggered the decision, the options considered, the chosen path, the decommissioning or redesign plan, and the governance actions required to safely wind down or restart the use case.

Retraining is the process of updating an AI model by training it on new or additional data to restore or improve its performance after drift, degradation, or changing business requirements.

Retrieval-Augmented Generation is a technique that enhances AI model responses by first retrieving relevant information from external knowledge sources -- databases, document repositories, knowledge bases -- and then using that information as context for generating more accurate, grounded answers.

A retrospective is a structured review session conducted after completing work to examine what went well, what went wrong, and what should change going forward.

Return on Investment (ROI) is a financial performance measure calculated as (net benefits minus costs) divided by costs, expressed as a percentage, used to evaluate the profitability of an AI transformation investment.

Reward hacking occurs when an AI agent learns to maximize its reward signal in unintended ways that do not align with the actual desired outcome.

Risk appetite is the overall level and types of risk that an organization is willing to accept in pursuit of its strategic objectives, set by the board of directors or equivalent governing body.

A formally approved document that defines the types and levels of AI-related risk the organization is willing to accept in pursuit of its AI ambition, covering operational, reputational, regulatory, and ethical risk dimensions.

A structured, evidence-based evaluation of the risks associated with each AI system or use case — covering technical, operational, ethical, regulatory, and reputational risk dimensions — with scored likelihood and impact ratings, existing control effectiveness ratings, and residual risk determinations.

A risk register is a comprehensive, living document that catalogs all identified AI-related risks for a transformation program or portfolio, recording each risk's description, probability assessment, potential impact, current mitigation measures, residual risk level, assigned owner, review frequency, and status.

A risk taxonomy is a structured classification system that organizes AI-specific risks into categories with defined severity levels, likelihood assessments, and mitigation strategies.

Risk-based classification is an approach to AI governance that applies different levels of regulatory requirements, oversight, and governance controls based on the potential risk of harm from the AI application.

A transformation roadmap is a strategic planning document that maps AI initiatives to timelines, resources, dependencies, milestones, and success criteria.

RPA is software that automates repetitive, rule-based tasks typically performed by humans, such as data entry, form filling, file moving, and report generation.

A comprehensive financial and strategic outcome report that quantifies the realized value of an AI initiative against its original value thesis — covering financial return, operational improvement, strategic enablement, and risk reduction — providing the evidence base for scaling, continuation, or retirement decisions.

Rollback reverts an AI system to a previous known-good state when the current version causes problems, requiring versioned model snapshots and automated reversion procedures.

SAFe (Scaled Agile Framework) is a comprehensive framework for implementing agile practices at enterprise scale, addressing the coordination challenges that arise when multiple agile teams must work together on large, complex initiatives.

Safety in AI means that systems are designed to operate reliably within their intended boundaries and fail gracefully when they encounter situations outside their training distribution.

Scalability is the ability to expand AI capabilities from individual successes to enterprise-wide deployment without proportional increases in effort or cost per deployment.

A formal record of the decision to scale, replicate, or extend an AI use case to additional business units, geographies, or processes — documenting the evidence basis, scaling approach, resource requirements, governance adaptations needed, and the named decision-makers who authorized the expansion.

Scope creep is the uncontrolled, incremental expansion of a project or engagement's requirements beyond its originally agreed boundaries, typically occurring through a series of individually reasonable additions that collectively consume disproportionate time, resources, and budget without corresponding value.

Scrum of scrums is a coordination mechanism where representatives from multiple agile teams meet regularly to share progress, surface cross-team dependencies, and resolve inter-team issues in large transformation programs with five or more concurrent workstreams.

Security by design is the principle of integrating security considerations into the architecture and design of AI systems from the earliest stages of development rather than adding security measures as an afterthought.

Self-sustaining capability is an organization's demonstrated ability to continue AI transformation, governance improvement, and capability development independently after external consultants depart.

Sensitivity analysis is a technique that tests how changes in key assumptions affect the outcomes of a business case, financial model, or risk assessment.

Sentiment analysis is a natural language processing technique that determines the emotional tone, opinion, or attitude expressed in text -- typically classified as positive, negative, or neutral, sometimes with finer-grained categories like anger, joy, or frustration.

A Service Level Agreement is a formal commitment between a service provider and consumer defining expected performance levels.

Shadow AI refers to AI tools, models, and AI-enabled applications that employees use within an organization without formal approval from IT, legal, risk management, or governance functions.

A structured catalogue of AI tools, models, and automated systems already in use across the organization that were deployed outside formal governance channels.

Shadow deployment (also called shadow mode) is a deployment pattern where a new AI model runs alongside the current production model, receiving the same real-world inputs but without its outputs being served to users or affecting business processes.

A showback model shows business units their AI resource consumption costs without billing them, creating awareness before full chargeback implementation.

A high-rigor credential requiring 50-80 hours of study that spans multiple COMPEL stages and includes a capstone project with peer review.

A formal, signed commitment from executive sponsors authorizing an AI transformation program and allocating the human, financial, and political capital required to sustain it.

A sprint is a fixed one-to-four-week period during which a transformation team commits to completing defined deliverables, providing the rhythmic heartbeat of the Produce stage through planning, execution, review, and retrospective.

The formal rules governing how credentials combine and contribute to higher-level credentials in the lattice.

A stage gate is a structured decision point between COMPEL lifecycle stages that verifies deliverables meet quality criteria before the organization advances.

A stakeholder is any individual, group, or organization that has an interest in or is affected by an AI transformation initiative.

Stakeholder alignment is the deliberate process of ensuring that all key stakeholders in an AI transformation program share a common understanding of objectives, success criteria, roles, governance mechanisms, risk tolerances, and expected outcomes before and during program execution.

A structured plan that identifies all stakeholders affected by the AI transformation program, maps their influence and interest, and defines the engagement approach, communication cadence, and escalation paths for each stakeholder group.

Stakeholder mapping identifies all individuals and organizations affected by or influential over AI transformation, documenting roles, interests, concerns, and relationships.

Standard Contractual Clauses (SCCs) are pre-approved contract terms established by the European Commission for transferring personal data from the EU to countries that do not have an adequacy decision.

A steering committee is a senior leadership body that provides strategic oversight, decision-making authority, cross-functional conflict resolution, and executive sponsorship for an AI transformation program.

Strategic risk encompasses threats to an organization's fundamental strategy, competitive position, or long-term viability, including the risk of falling behind competitors in AI capability, making wrong technology platform bets, failing to attract and retain AI talent, or being disrupted by AI-native competitors.

Structured data is data organized in a predefined format with rows and columns, such as spreadsheets, database tables, ERP records, and CRM entries.

Summative assessment is a final evaluation conducted at the conclusion of a learning program or certification process to determine whether participants have achieved the required learning objectives and competency levels.

Supervised learning is the most widely deployed machine learning paradigm in enterprises.

Supply chain AI governance manages risks and accountability across organizational boundaries through vendor relationships, third-party models, and partner integrations.

Synthetic data is artificially generated data that mimics the statistical properties of real data but does not contain actual individual records.

Systems thinking is an approach that views AI initiatives not as isolated technology projects but as interventions in a complex organizational system where changes ripple through upstream and downstream workflows, employee roles, customer interactions, data flows, and governance processes.

Talent strategy is the comprehensive plan for building and sustaining the human capabilities needed for AI transformation, encompassing workforce assessment, role definition, hiring, reskilling and upskilling programs, career pathway design, retention mechanisms, and organizational development.

The technical CE cap is the maximum percentage of Continuing Education credits that can be earned from purely technical activities (such as completing partner bootcamps, passing technical assessments, or attending technical conferences) toward a COMPEL certification renewal.

Technical debt is the accumulated cost of shortcuts, workarounds, and deferred maintenance in technology systems that become increasingly expensive to address over time.

Technical feasibility is an assessment of whether a proposed AI solution can be practically built and deployed given current technology capabilities, data availability, infrastructure, organizational skills, and time constraints.

Telemetry is the automated collection of operational data about AI systems including performance metrics, resource consumption, model behavior, and user interaction patterns providing raw signals for monitoring and alerting.

The documented specification and implemented configuration of all monitoring, logging, alerting, and observability instrumentation for an AI system — covering model performance metrics, data drift indicators, governance control effectiveness signals, and operational health metrics.

Thought leadership creates and shares original insights advancing AI transformation through publications, presentations, and community engagement.

The three lines of defense is a widely adopted risk governance model that distributes risk management responsibilities across three organizational levels: the first line (operational management and AI teams) owns and manages risks directly in their daily work; the second line (risk management and compliance functions) provides oversight, policies, and guidance; and the third line (internal audit) provides independent assurance that the first and second lines are functioning effectively.

TOGAF (The Open Group Architecture Framework) is a widely used enterprise architecture methodology that provides a structured approach for designing, planning, implementing, and governing enterprise information technology architecture.

A token is the basic unit of text that a language model processes, roughly corresponding to a word or word fragment (typically 3-4 characters in English).

The token cost multiplier is the factor by which AI token consumption increases in multi-agent systems compared to single-model interactions, reflecting the additional tokens consumed by inter-agent communication, extended reasoning chains, tool call sequences, error recovery conversations, and coordination overhead.

Token economics is the analysis, budgeting, and optimization of costs associated with AI language model usage, where pricing is based on the number of tokens (text units, typically representing about four characters) processed as input and generated as output.

Tool call authorization controls which tools and APIs an AI agent may access, with what parameters and approval requirements.

Total Cost of Ownership is a comprehensive financial analysis that captures the complete cost of an AI system over its entire lifecycle, including initial development, infrastructure, data acquisition, ongoing maintenance, model retraining, monitoring, governance compliance, talent, vendor fees, and eventual decommissioning.

A TPU is a custom-designed processor created by Google specifically for neural network workloads, available through Google Cloud Platform.

A structured plan that defines the training content, delivery methods, timing, audience segmentation, success criteria, and adoption measurement approach for every user group affected by an AI system deployment.

Training data is the dataset used to teach a machine learning model the patterns it needs to make predictions or generate outputs.

A transformation crisis is a critical event threatening AI program success including executive departure, budget cuts, major failures, or public controversy.

Transformation Enablers are three cross-cutting layers in the COMPEL framework -- Value Realization, Operational Readiness, and Agent Governance -- that operate horizontally across all six lifecycle stages.

A transformation portfolio is the collection of AI programs and initiatives managed together to achieve enterprise strategic objectives, balanced across risk, time horizons, pillar coverage, and capability dependencies.

A transformation roadmap is a strategic, time-sequenced plan that organizes AI transformation initiatives across the four COMPEL pillars, showing what will be done, in what order, with what resources, by when, and with what expected outcomes.

A transformation sprint is a two-week time-boxed work period within the COMPEL Produce stage that delivers concrete outcomes across multiple pillars simultaneously.

The transformer is the neural network architecture that powers modern large language models and many other state-of-the-art AI systems.

Transparency in AI governance is the principle that organizations should openly communicate about their use of AI, how AI systems make decisions, what data they use, what their limitations are, and what governance mechanisms are in place.

The trust dividend is the compound return that organizations earn from investing in responsible AI practices, accruing across multiple dimensions of stakeholder relationships.

Trustworthy AI describes AI systems that are lawful (complying with all applicable regulations), ethical (adhering to moral principles and values), and robust (technically reliable, safe, and secure).

Uncertainty estimation encompasses the techniques and methods for quantifying how confident an AI model is in its individual predictions, enabling downstream systems and users to make informed decisions about when to trust AI outputs and when to escalate to human judgment or alternative decision processes.

Uncertainty quantification encompasses methods for measuring and communicating how confident an AI model is in its predictions.

Unstructured data is data that does not follow a predefined format, including text documents, images, audio recordings, video files, emails, chat transcripts, and social media content.

Unsupervised learning is a machine learning approach that discovers hidden patterns and structures in data without pre-labeled examples.

Uplift modeling estimates the incremental impact of interventions on individual outcomes, identifying who benefits most versus those unaffected regardless.

In AI transformation, a use case is a specific application of AI to a defined business problem with measurable outcomes, identifiable stakeholders, and quantifiable resource requirements.

Use case intake is the structured process of collecting, documenting, evaluating, and prioritizing proposed AI use cases from across the organization through a standardized submission and review workflow.

A use case portfolio is a deliberately balanced collection of AI initiatives designed to achieve strategic outcomes while managing risk across a COMPEL cycle.

A structured prioritization tool that maps candidate AI use cases across value potential, feasibility, risk, and strategic alignment dimensions to produce a ranked, resource-constrained portfolio for the transformation program.

User Acceptance Testing (UAT) is the final validation phase before an AI system goes into production, where actual end users (not the development team) evaluate whether the system meets their operational needs, integrates into their workflows, and produces acceptable results in realistic working conditions.

A validation framework verifies AI systems across accuracy, fairness, robustness, security, and explainability, defining methods, metrics, and thresholds before production.

Value attribution determines how much observed business outcome can be credited to AI transformation versus other factors.

The end-to-end process of defining, tracking, and verifying the business value delivered by AI initiatives — from initial value thesis through baseline measurement, deployment, post-deployment review, and ongoing benefit tracking.

The value realization framework is a structured methodology for defining, measuring, tracking, and verifying business value delivery from AI transformation initiatives throughout the entire COMPEL lifecycle.

A value thesis is a testable hypothesis articulating the causal logic connecting an AI initiative to expected business outcomes.

A living register that documents the specific value thesis for each AI use case in the portfolio — including the problem statement, value hypothesis, measurement approach, baseline, and expected outcomes.

A vector database is a specialized database designed to store and efficiently search high-dimensional numerical representations (embeddings) of data.

Vendor due diligence is the structured investigation of an AI vendor's or partner's capabilities, security practices, data handling procedures, compliance posture, financial stability, support quality, and contractual terms before entering a business relationship or deploying their technology.

A vendor ecosystem is the network of external technology providers, service partners, cloud platforms, specialized AI tool vendors, and consulting firms that an organization relies on for AI capabilities, infrastructure, and expertise.

A vendor risk assessment evaluates the governance risks introduced by third-party AI components that an organization depends on, including foundation model providers, MLOps platforms, data services, labeling providers, and AI-as-a-service offerings.

Version control is the practice of tracking and managing changes to code, data, models, and configuration files over time, maintaining a complete history of what changed, when, who made the change, and why.

Vulnerability scanning is the automated process of testing AI systems, supporting infrastructure, and related software for known security weaknesses, misconfigurations, and exploitable flaws.

Warm start is a training technique where an AI model begins its learning process using the weights and parameters from a previously trained model rather than starting from random values, significantly reducing training time and computational cost while often improving final performance.

Waterfall is a linear project management approach where phases (requirements, design, implementation, testing, deployment) are completed sequentially from start to finish.

Weight decay is a regularization technique used during AI model training that adds a penalty term proportional to the magnitude of model weights, discouraging the model from relying too heavily on any single feature and promoting simpler, more generalizable models.

Whistleblower protection encompasses policies and mechanisms that protect individuals who report AI-related concerns, ethical violations, governance failures, or safety risks from retaliation, career consequences, or social punishment.

Workflow orchestration is the automated coordination of complex, multi-step processes that involve multiple systems, services, human participants, or AI agents, managing the sequence of steps, parallel execution paths, conditional branching, error handling, retry logic, and completion tracking.

The deliberate restructuring of business processes to embed AI capabilities at optimal human-AI handoff points, rather than overlaying AI onto existing manual workflows.

Structured documentation of the before and after states of each business process where an AI system is being embedded — covering process maps, human-AI handoff points, role changes, exception handling paths, and the rationale for each design decision.

Workforce AI capability transformation is the systematic process of assessing, redesigning, and developing an organization's workforce capabilities to operate effectively in an AI-augmented environment.

A structured plan that defines the AI skills, roles, and capabilities the organization needs to execute its transformation program, assesses current workforce gaps, and specifies the training, hiring, and capability-building interventions required to close those gaps by the time they are needed.

Workforce redesign is the process of analyzing and restructuring jobs at the task level to determine which tasks are best automated by AI, which are augmented by AI, and which remain fully human.

Workforce strategy plans how human resources evolve through assessment, reskilling, hiring, restructuring, and retention for AI transformation.

Workforce transformation is the strategic process of developing new skills, redesigning roles, restructuring teams, and evolving organizational culture to enable effective human-AI collaboration across the enterprise.

X-risk refers to the theoretical existential risk that sufficiently advanced AI systems could pose catastrophic or irreversible harm to humanity or civilization.

XAI is the abbreviated term for Explainable Artificial Intelligence, the field focused on making AI systems' reasoning and decision-making processes transparent and interpretable to humans.

XAI techniques are specific methods making AI decisions interpretable including SHAP values, LIME, attention visualization, feature importance, and counterfactual explanations.

XGBoost (eXtreme Gradient Boosting) is a highly efficient and widely used machine learning algorithm that builds predictions by combining many small decision trees in sequence, with each tree learning from the errors of the previous ones.

XML is a structured data format widely used for storing, transmitting, and exchanging data between different software systems in a human-readable and machine-parseable format.

XML and related formats provide standardized ways to structure, share, and validate data between systems and organizations.

YAML (YAML Ain't Markup Language) is a human-readable data serialization format commonly used for configuration files in AI/ML pipelines, infrastructure-as-code definitions, CI/CD pipeline specifications, and deployment configurations.

YAML (YAML Ain't Markup Language) is a human-readable data serialization format commonly used to define configuration settings, pipeline specifications, infrastructure definitions, and deployment parameters for AI systems and their supporting infrastructure.

Year-over-year maturity progression measures how an organization's AI maturity scores change across annual assessment cycles.

Year-over-year (YoY) metrics compare performance data from the same period in consecutive years, providing a normalized view of long-term AI transformation progress that accounts for seasonal variations, cyclical patterns, and short-term fluctuations.

Yield management dynamically adjusts resource allocation based on demand patterns to maximize value from scarce AI infrastructure capacity, applying to GPU utilization, inference versus training balance, and workload scheduling.

Yield optimization uses AI to maximize the output, efficiency, or return from a process -- such as manufacturing yield (reducing waste and defects), agricultural yield (optimizing crop production), advertising yield (maximizing revenue per impression), or financial yield (optimizing investment returns).

A z-score is a statistical measurement describing how many standard deviations a data point is from the mean (average) of its dataset.

A zero-day vulnerability is a software security flaw that is unknown to the software vendor and therefore has no available patch or fix at the time of discovery.

Zero-shot learning is the ability of an AI model to perform tasks it was not explicitly trained or fine-tuned to do, leveraging general knowledge and reasoning capabilities acquired during pre-training.

Zero-trust architecture is a security framework built on the principle that no user, device, system, or AI agent should be trusted by default, regardless of whether they are inside or outside the network perimeter.

The zone of proximal development (ZPD), originally theorized by Lev Vygotsky, describes the gap between what a learner can accomplish independently and what they can achieve with appropriate guidance and support.